mirror of
https://github.com/Relintai/pmlpp.git
synced 2024-11-08 13:12:09 +01:00
2799 lines
88 KiB
C++
2799 lines
88 KiB
C++
/*************************************************************************/
|
|
/* activation.cpp */
|
|
/*************************************************************************/
|
|
/* This file is part of: */
|
|
/* PMLPP Machine Learning Library */
|
|
/* https://github.com/Relintai/pmlpp */
|
|
/*************************************************************************/
|
|
/* Copyright (c) 2023-present Péter Magyar. */
|
|
/* Copyright (c) 2022-2023 Marc Melikyan */
|
|
/* */
|
|
/* Permission is hereby granted, free of charge, to any person obtaining */
|
|
/* a copy of this software and associated documentation files (the */
|
|
/* "Software"), to deal in the Software without restriction, including */
|
|
/* without limitation the rights to use, copy, modify, merge, publish, */
|
|
/* distribute, sublicense, and/or sell copies of the Software, and to */
|
|
/* permit persons to whom the Software is furnished to do so, subject to */
|
|
/* the following conditions: */
|
|
/* */
|
|
/* The above copyright notice and this permission notice shall be */
|
|
/* included in all copies or substantial portions of the Software. */
|
|
/* */
|
|
/* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, */
|
|
/* EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF */
|
|
/* MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.*/
|
|
/* IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY */
|
|
/* CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, */
|
|
/* TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE */
|
|
/* SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */
|
|
/*************************************************************************/
|
|
|
|
#include "activation.h"
|
|
#include "../lin_alg/lin_alg.h"
|
|
|
|
#include <algorithm>
|
|
#include <cmath>
|
|
#include <iostream>
|
|
|
|
MLPPActivation::RealActivationFunctionPointer MLPPActivation::get_activation_function_ptr_real(const ActivationFunction func, const bool deriv) {
|
|
if (deriv) {
|
|
return get_activation_function_ptr_normal_real(func);
|
|
} else {
|
|
return get_activation_function_ptr_deriv_real(func);
|
|
}
|
|
}
|
|
MLPPActivation::VectorActivationFunctionPointer MLPPActivation::get_activation_function_ptr_vector(const ActivationFunction func, const bool deriv) {
|
|
if (deriv) {
|
|
return get_activation_function_ptr_normal_vector(func);
|
|
} else {
|
|
return get_activation_function_ptr_deriv_vector(func);
|
|
}
|
|
}
|
|
MLPPActivation::MatrixActivationFunctionPointer MLPPActivation::get_activation_function_ptr_matrix(const ActivationFunction func, const bool deriv) {
|
|
if (deriv) {
|
|
return get_activation_function_ptr_normal_matrix(func);
|
|
} else {
|
|
return get_activation_function_ptr_deriv_matrix(func);
|
|
}
|
|
}
|
|
|
|
MLPPActivation::RealActivationFunctionPointer MLPPActivation::get_activation_function_ptr_normal_real(const ActivationFunction func) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return &MLPPActivation::linear_normr;
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return &MLPPActivation::sigmoid_normr;
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return &MLPPActivation::swish_normr;
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return &MLPPActivation::mish_normr;
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return &MLPPActivation::sinc_normr;
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return &MLPPActivation::softmax_normr;
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return &MLPPActivation::softplus_normr;
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return &MLPPActivation::softsign_normr;
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return &MLPPActivation::adj_softmax_normr;
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return &MLPPActivation::cloglog_normr;
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return &MLPPActivation::logit_normr;
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return &MLPPActivation::gaussian_cdf_normr;
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return &MLPPActivation::relu_normr;
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return &MLPPActivation::gelu_normr;
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return &MLPPActivation::sign_normr;
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return &MLPPActivation::unit_step_normr;
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return &MLPPActivation::sinh_normr;
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return &MLPPActivation::cosh_normr;
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return &MLPPActivation::tanh_normr;
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return &MLPPActivation::csch_normr;
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return &MLPPActivation::sech_normr;
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return &MLPPActivation::coth_normr;
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return &MLPPActivation::arsinh_normr;
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return &MLPPActivation::arcosh_normr;
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return &MLPPActivation::artanh_normr;
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return &MLPPActivation::arcsch_normr;
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return &MLPPActivation::arsech_normr;
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return &MLPPActivation::arcoth_normr;
|
|
default:
|
|
return NULL;
|
|
}
|
|
}
|
|
MLPPActivation::VectorActivationFunctionPointer MLPPActivation::get_activation_function_ptr_normal_vector(const ActivationFunction func) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return &MLPPActivation::linear_normv;
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return &MLPPActivation::sigmoid_normv;
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return &MLPPActivation::swish_normv;
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return &MLPPActivation::mish_normv;
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return &MLPPActivation::sinc_normv;
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return &MLPPActivation::softmax_normv;
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return &MLPPActivation::softplus_normv;
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return &MLPPActivation::softsign_normv;
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return &MLPPActivation::adj_softmax_normv;
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return &MLPPActivation::cloglog_normv;
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return &MLPPActivation::logit_normv;
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return &MLPPActivation::gaussian_cdf_normv;
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return &MLPPActivation::relu_normv;
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return &MLPPActivation::gelu_normv;
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return &MLPPActivation::sign_normv;
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return &MLPPActivation::unit_step_normv;
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return &MLPPActivation::sinh_normv;
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return &MLPPActivation::cosh_normv;
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return &MLPPActivation::tanh_normv;
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return &MLPPActivation::csch_normv;
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return &MLPPActivation::sech_normv;
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return &MLPPActivation::coth_normv;
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return &MLPPActivation::arsinh_normv;
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return &MLPPActivation::arcosh_normv;
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return &MLPPActivation::artanh_normv;
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return &MLPPActivation::arcsch_normv;
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return &MLPPActivation::arsech_normv;
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return &MLPPActivation::arcoth_normv;
|
|
default:
|
|
return NULL;
|
|
}
|
|
}
|
|
MLPPActivation::MatrixActivationFunctionPointer MLPPActivation::get_activation_function_ptr_normal_matrix(const ActivationFunction func) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return &MLPPActivation::linear_normm;
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return &MLPPActivation::sigmoid_normm;
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return &MLPPActivation::swish_normm;
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return &MLPPActivation::mish_normm;
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return &MLPPActivation::sinc_normm;
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return &MLPPActivation::softmax_normm;
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return &MLPPActivation::softplus_normm;
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return &MLPPActivation::softsign_normm;
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return &MLPPActivation::adj_softmax_normm;
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return &MLPPActivation::cloglog_normm;
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return &MLPPActivation::logit_normm;
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return &MLPPActivation::gaussian_cdf_normm;
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return &MLPPActivation::relu_normm;
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return &MLPPActivation::gelu_normm;
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return &MLPPActivation::sign_normm;
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return &MLPPActivation::unit_step_normm;
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return &MLPPActivation::sinh_normm;
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return &MLPPActivation::cosh_normm;
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return &MLPPActivation::tanh_normm;
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return &MLPPActivation::csch_normm;
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return &MLPPActivation::sech_normm;
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return &MLPPActivation::coth_normm;
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return &MLPPActivation::arsinh_normm;
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return &MLPPActivation::arcosh_normm;
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return &MLPPActivation::artanh_normm;
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return &MLPPActivation::arcsch_normm;
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return &MLPPActivation::arsech_normm;
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return &MLPPActivation::arcoth_normm;
|
|
default:
|
|
return NULL;
|
|
}
|
|
}
|
|
|
|
MLPPActivation::RealActivationFunctionPointer MLPPActivation::get_activation_function_ptr_deriv_real(const ActivationFunction func) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return &MLPPActivation::linear_normr;
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return &MLPPActivation::sigmoid_normr;
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return &MLPPActivation::swish_normr;
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return &MLPPActivation::mish_normr;
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return &MLPPActivation::sinc_normr;
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return &MLPPActivation::softmax_normr;
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return &MLPPActivation::softplus_normr;
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return &MLPPActivation::softsign_normr;
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return &MLPPActivation::adj_softmax_normr;
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return &MLPPActivation::cloglog_normr;
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return &MLPPActivation::logit_normr;
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return &MLPPActivation::gaussian_cdf_normr;
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return &MLPPActivation::relu_normr;
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return &MLPPActivation::gelu_normr;
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return &MLPPActivation::sign_normr;
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return &MLPPActivation::unit_step_normr;
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return &MLPPActivation::sinh_normr;
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return &MLPPActivation::cosh_normr;
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return &MLPPActivation::tanh_normr;
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return &MLPPActivation::csch_normr;
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return &MLPPActivation::sech_normr;
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return &MLPPActivation::coth_normr;
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return &MLPPActivation::arsinh_normr;
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return &MLPPActivation::arcosh_normr;
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return &MLPPActivation::artanh_normr;
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return &MLPPActivation::arcsch_normr;
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return &MLPPActivation::arsech_normr;
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return &MLPPActivation::arcoth_normr;
|
|
default:
|
|
return NULL;
|
|
}
|
|
}
|
|
MLPPActivation::VectorActivationFunctionPointer MLPPActivation::get_activation_function_ptr_deriv_vector(const ActivationFunction func) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return &MLPPActivation::linear_derivv;
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return &MLPPActivation::sigmoid_derivv;
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return &MLPPActivation::swish_derivv;
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return &MLPPActivation::mish_derivv;
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return &MLPPActivation::sinc_derivv;
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return &MLPPActivation::softmax_derivv;
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return &MLPPActivation::softplus_derivv;
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return &MLPPActivation::softsign_derivv;
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return &MLPPActivation::adj_softmax_derivv;
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return &MLPPActivation::cloglog_derivv;
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return &MLPPActivation::logit_derivv;
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return &MLPPActivation::gaussian_cdf_derivv;
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return &MLPPActivation::relu_derivv;
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return &MLPPActivation::gelu_derivv;
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return &MLPPActivation::sign_derivv;
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return &MLPPActivation::unit_step_derivv;
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return &MLPPActivation::sinh_derivv;
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return &MLPPActivation::cosh_derivv;
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return &MLPPActivation::tanh_derivv;
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return &MLPPActivation::csch_derivv;
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return &MLPPActivation::sech_derivv;
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return &MLPPActivation::coth_derivv;
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return &MLPPActivation::arsinh_derivv;
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return &MLPPActivation::arcosh_derivv;
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return &MLPPActivation::artanh_derivv;
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return &MLPPActivation::arcsch_derivv;
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return &MLPPActivation::arsech_derivv;
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return &MLPPActivation::arcoth_derivv;
|
|
default:
|
|
return NULL;
|
|
}
|
|
}
|
|
MLPPActivation::MatrixActivationFunctionPointer MLPPActivation::get_activation_function_ptr_deriv_matrix(const ActivationFunction func) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return &MLPPActivation::linear_derivm;
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return &MLPPActivation::sigmoid_derivm;
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return &MLPPActivation::swish_derivm;
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return &MLPPActivation::mish_derivm;
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return &MLPPActivation::sinc_derivm;
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return &MLPPActivation::softmax_derivm;
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return &MLPPActivation::softplus_derivm;
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return &MLPPActivation::softsign_derivm;
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return &MLPPActivation::adj_softmax_derivm;
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return &MLPPActivation::cloglog_derivm;
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return &MLPPActivation::logit_derivm;
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return &MLPPActivation::gaussian_cdf_derivm;
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return &MLPPActivation::relu_derivm;
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return &MLPPActivation::gelu_derivm;
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return &MLPPActivation::sign_derivm;
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return &MLPPActivation::unit_step_derivm;
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return &MLPPActivation::sinh_derivm;
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return &MLPPActivation::cosh_derivm;
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return &MLPPActivation::tanh_derivm;
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return &MLPPActivation::csch_derivm;
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return &MLPPActivation::sech_derivm;
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return &MLPPActivation::coth_derivm;
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return &MLPPActivation::arsinh_derivm;
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return &MLPPActivation::arcosh_derivm;
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return &MLPPActivation::artanh_derivm;
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return &MLPPActivation::arcsch_derivm;
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return &MLPPActivation::arsech_derivm;
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return &MLPPActivation::arcoth_derivm;
|
|
default:
|
|
return NULL;
|
|
}
|
|
}
|
|
|
|
real_t MLPPActivation::run_activation_real(const ActivationFunction func, const real_t z, const bool deriv) {
|
|
if (deriv) {
|
|
return run_activation_norm_real(func, z);
|
|
} else {
|
|
return run_activation_deriv_real(func, z);
|
|
}
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::run_activation_vector(const ActivationFunction func, const Ref<MLPPVector> &z, const bool deriv) {
|
|
if (deriv) {
|
|
return run_activation_norm_vector(func, z);
|
|
} else {
|
|
return run_activation_deriv_vector(func, z);
|
|
}
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::run_activation_matrix(const ActivationFunction func, const Ref<MLPPMatrix> &z, const bool deriv) {
|
|
if (deriv) {
|
|
return run_activation_norm_matrix(func, z);
|
|
} else {
|
|
return run_activation_deriv_matrix(func, z);
|
|
}
|
|
}
|
|
|
|
real_t MLPPActivation::run_activation_norm_real(const ActivationFunction func, const real_t z) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return linear_normr(z);
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return sigmoid_normr(z);
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return swish_normr(z);
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return mish_normr(z);
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return sinc_normr(z);
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return softmax_normr(z);
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return softplus_normr(z);
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return softsign_normr(z);
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return adj_softmax_normr(z);
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return cloglog_normr(z);
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return logit_normr(z);
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return gaussian_cdf_normr(z);
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return relu_normr(z);
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return gelu_normr(z);
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return sign_normr(z);
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return unit_step_normr(z);
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return sinh_normr(z);
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return cosh_normr(z);
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return tanh_normr(z);
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return csch_normr(z);
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return sech_normr(z);
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return coth_normr(z);
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return arsinh_normr(z);
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return arcosh_normr(z);
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return artanh_normr(z);
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return arcsch_normr(z);
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return arsech_normr(z);
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return arcoth_normr(z);
|
|
default:
|
|
ERR_FAIL_V(0);
|
|
}
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::run_activation_norm_vector(const ActivationFunction func, const Ref<MLPPVector> &z) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return linear_normv(z);
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return sigmoid_normv(z);
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return swish_normv(z);
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return mish_normv(z);
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return sinc_normv(z);
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return softmax_normv(z);
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return softplus_normv(z);
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return softsign_normv(z);
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return adj_softmax_normv(z);
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return cloglog_normv(z);
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return logit_normv(z);
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return gaussian_cdf_normv(z);
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return relu_normv(z);
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return gelu_normv(z);
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return sign_normv(z);
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return unit_step_normv(z);
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return sinh_normv(z);
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return cosh_normv(z);
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return tanh_normv(z);
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return csch_normv(z);
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return sech_normv(z);
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return coth_normv(z);
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return arsinh_normv(z);
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return arcosh_normv(z);
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return artanh_normv(z);
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return arcsch_normv(z);
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return arsech_normv(z);
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return arcoth_normv(z);
|
|
default:
|
|
ERR_FAIL_V(Ref<MLPPVector>());
|
|
}
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::run_activation_norm_matrix(const ActivationFunction func, const Ref<MLPPMatrix> &z) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return linear_normm(z);
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return sigmoid_normm(z);
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return swish_normm(z);
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return mish_normm(z);
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return sinc_normm(z);
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return softmax_normm(z);
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return softplus_normm(z);
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return softsign_normm(z);
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return adj_softmax_normm(z);
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return cloglog_normm(z);
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return logit_normm(z);
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return gaussian_cdf_normm(z);
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return relu_normm(z);
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return gelu_normm(z);
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return sign_normm(z);
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return unit_step_normm(z);
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return sinh_normm(z);
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return cosh_normm(z);
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return tanh_normm(z);
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return csch_normm(z);
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return sech_normm(z);
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return coth_normm(z);
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return arsinh_normm(z);
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return arcosh_normm(z);
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return artanh_normm(z);
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return arcsch_normm(z);
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return arsech_normm(z);
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return arcoth_normm(z);
|
|
default:
|
|
ERR_FAIL_V(Ref<MLPPMatrix>());
|
|
}
|
|
}
|
|
|
|
real_t MLPPActivation::run_activation_deriv_real(const ActivationFunction func, const real_t z) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return linear_normr(z);
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return sigmoid_normr(z);
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return swish_normr(z);
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return mish_normr(z);
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return sinc_normr(z);
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return softmax_normr(z);
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return softplus_normr(z);
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return softsign_normr(z);
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return adj_softmax_normr(z);
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return cloglog_normr(z);
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return logit_normr(z);
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return gaussian_cdf_normr(z);
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return relu_normr(z);
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return gelu_normr(z);
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return sign_normr(z);
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return unit_step_normr(z);
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return sinh_normr(z);
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return cosh_normr(z);
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return tanh_normr(z);
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return csch_normr(z);
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return sech_normr(z);
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return coth_normr(z);
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return arsinh_normr(z);
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return arcosh_normr(z);
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return artanh_normr(z);
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return arcsch_normr(z);
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return arsech_normr(z);
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return arcoth_normr(z);
|
|
default:
|
|
ERR_FAIL_V(0);
|
|
}
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::run_activation_deriv_vector(const ActivationFunction func, const Ref<MLPPVector> &z) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return linear_derivv(z);
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return sigmoid_derivv(z);
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return swish_derivv(z);
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return mish_derivv(z);
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return sinc_derivv(z);
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return softmax_derivv(z);
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return softplus_derivv(z);
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return softsign_derivv(z);
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return adj_softmax_derivv(z);
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return cloglog_derivv(z);
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return logit_derivv(z);
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return gaussian_cdf_derivv(z);
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return relu_derivv(z);
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return gelu_derivv(z);
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return sign_derivv(z);
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return unit_step_derivv(z);
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return sinh_derivv(z);
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return cosh_derivv(z);
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return tanh_derivv(z);
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return csch_derivv(z);
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return sech_derivv(z);
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return coth_derivv(z);
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return arsinh_derivv(z);
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return arcosh_derivv(z);
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return artanh_derivv(z);
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return arcsch_derivv(z);
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return arsech_derivv(z);
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return arcoth_derivv(z);
|
|
default:
|
|
ERR_FAIL_V(Ref<MLPPVector>());
|
|
}
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::run_activation_deriv_matrix(const ActivationFunction func, const Ref<MLPPMatrix> &z) {
|
|
switch (func) {
|
|
case ACTIVATION_FUNCTION_LINEAR:
|
|
return linear_derivm(z);
|
|
case ACTIVATION_FUNCTION_SIGMOID:
|
|
return sigmoid_derivm(z);
|
|
case ACTIVATION_FUNCTION_SWISH:
|
|
return swish_derivm(z);
|
|
case ACTIVATION_FUNCTION_MISH:
|
|
return mish_derivm(z);
|
|
case ACTIVATION_FUNCTION_SIN_C:
|
|
return sinc_derivm(z);
|
|
case ACTIVATION_FUNCTION_SOFTMAX:
|
|
return softmax_derivm(z);
|
|
case ACTIVATION_FUNCTION_SOFTPLUS:
|
|
return softplus_derivm(z);
|
|
case ACTIVATION_FUNCTION_SOFTSIGN:
|
|
return softsign_derivm(z);
|
|
case ACTIVATION_FUNCTION_ADJ_SOFTMAX:
|
|
return adj_softmax_derivm(z);
|
|
case ACTIVATION_FUNCTION_C_LOG_LOG:
|
|
return cloglog_derivm(z);
|
|
case ACTIVATION_FUNCTION_LOGIT:
|
|
return logit_derivm(z);
|
|
case ACTIVATION_FUNCTION_GAUSSIAN_CDF:
|
|
return gaussian_cdf_derivm(z);
|
|
case ACTIVATION_FUNCTION_RELU:
|
|
return relu_derivm(z);
|
|
case ACTIVATION_FUNCTION_GELU:
|
|
return gelu_derivm(z);
|
|
case ACTIVATION_FUNCTION_SIGN:
|
|
return sign_derivm(z);
|
|
case ACTIVATION_FUNCTION_UNIT_STEP:
|
|
return unit_step_derivm(z);
|
|
case ACTIVATION_FUNCTION_SINH:
|
|
return sinh_derivm(z);
|
|
case ACTIVATION_FUNCTION_COSH:
|
|
return cosh_derivm(z);
|
|
case ACTIVATION_FUNCTION_TANH:
|
|
return tanh_derivm(z);
|
|
case ACTIVATION_FUNCTION_CSCH:
|
|
return csch_derivm(z);
|
|
case ACTIVATION_FUNCTION_SECH:
|
|
return sech_derivm(z);
|
|
case ACTIVATION_FUNCTION_COTH:
|
|
return coth_derivm(z);
|
|
case ACTIVATION_FUNCTION_ARSINH:
|
|
return arsinh_derivm(z);
|
|
case ACTIVATION_FUNCTION_ARCOSH:
|
|
return arcosh_derivm(z);
|
|
case ACTIVATION_FUNCTION_ARTANH:
|
|
return artanh_derivm(z);
|
|
case ACTIVATION_FUNCTION_ARCSCH:
|
|
return arcsch_derivm(z);
|
|
case ACTIVATION_FUNCTION_ARSECH:
|
|
return arsech_derivm(z);
|
|
case ACTIVATION_FUNCTION_ARCOTH:
|
|
return arcoth_derivm(z);
|
|
default:
|
|
ERR_FAIL_V(Ref<MLPPMatrix>());
|
|
}
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::activationr(const Ref<MLPPVector> &z, real_t (*function)(real_t)) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
|
|
int size = z->size();
|
|
|
|
a->resize(size);
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < size; ++i) {
|
|
a_ptr[i] = function(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
//ACTIVATION FUNCTIONS
|
|
|
|
//LINEAR
|
|
real_t MLPPActivation::linear_normr(real_t z) {
|
|
return z;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::linear_normv(const Ref<MLPPVector> &z) {
|
|
return z->duplicate_fast();
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::linear_normm(const Ref<MLPPMatrix> &z) {
|
|
return z->duplicate_fast();
|
|
}
|
|
|
|
real_t MLPPActivation::linear_derivr(real_t z) {
|
|
return 1;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::linear_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
return alg.onevecnv(z->size());
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::linear_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
return alg.onematnm(z->size().y, z->size().x);
|
|
}
|
|
|
|
//SIGMOID
|
|
real_t MLPPActivation::sigmoid_normr(real_t z) {
|
|
return 1 / (1 + exp(-z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::sigmoid_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
return alg.division_element_wisenv(alg.onevecnv(z->size()), alg.additionnv(alg.onevecnv(z->size()), alg.expnv(alg.scalar_multiplynv(-1, z))));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sigmoid_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
return alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), alg.additionnm(alg.onematnm(z->size().y, z->size().x), alg.expnm(alg.scalar_multiplynm(-1, z))));
|
|
}
|
|
|
|
real_t MLPPActivation::sigmoid_derivr(real_t z) {
|
|
real_t sig_norm = sigmoid_normr(z);
|
|
|
|
return sig_norm * (1 - sig_norm);
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::sigmoid_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
Ref<MLPPVector> sig_norm = sigmoid_normv(z);
|
|
|
|
return alg.subtractionnv(sig_norm, alg.hadamard_productnv(sig_norm, sig_norm));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sigmoid_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
Ref<MLPPMatrix> sig_norm = sigmoid_normm(z);
|
|
|
|
return alg.subtractionnm(sig_norm, alg.hadamard_productnm(sig_norm, sig_norm));
|
|
}
|
|
|
|
//SOFTMAX
|
|
|
|
real_t MLPPActivation::softmax_normr(real_t z) {
|
|
return z;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::softmax_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
int z_size = z->size();
|
|
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z_size);
|
|
|
|
Ref<MLPPVector> exp_z = alg.expnv(z);
|
|
real_t sum = 0;
|
|
|
|
const real_t *exp_z_ptr = exp_z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
sum += exp_z_ptr[i];
|
|
}
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = exp_z_ptr[i] / sum;
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::softmax_normm(const Ref<MLPPMatrix> &z) {
|
|
Size2i z_size = z->size();
|
|
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z_size);
|
|
|
|
Ref<MLPPVector> row_tmp;
|
|
row_tmp.instance();
|
|
row_tmp->resize(z_size.x);
|
|
|
|
for (int i = 0; i < z_size.y; ++i) {
|
|
z->row_get_into_mlpp_vector(i, row_tmp);
|
|
|
|
Ref<MLPPVector> sfn = softmax_normv(row_tmp);
|
|
|
|
a->row_set_mlpp_vector(i, sfn);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
real_t MLPPActivation::softmax_derivr(real_t z) {
|
|
return z;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::softmax_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
int z_size = z->size();
|
|
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z_size);
|
|
|
|
Ref<MLPPVector> exp_z = alg.expnv(z);
|
|
real_t sum = 0;
|
|
|
|
const real_t *exp_z_ptr = exp_z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
sum += exp_z_ptr[i];
|
|
}
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = exp_z_ptr[i] / sum;
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::softmax_derivm(const Ref<MLPPMatrix> &z) {
|
|
Size2i z_size = z->size();
|
|
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z_size);
|
|
|
|
Ref<MLPPVector> row_tmp;
|
|
row_tmp.instance();
|
|
row_tmp->resize(z_size.x);
|
|
|
|
for (int i = 0; i < z_size.y; ++i) {
|
|
z->row_get_into_mlpp_vector(i, row_tmp);
|
|
|
|
Ref<MLPPVector> sfn = softmax_derivm(z);
|
|
|
|
a->row_set_mlpp_vector(i, sfn);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
//ADJ_SOFTMAX
|
|
|
|
real_t MLPPActivation::adj_softmax_normr(real_t z) {
|
|
return 0;
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::adj_softmax_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
int size = z->size();
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t c = -Math_INF;
|
|
|
|
for (int i = 0; i < size; ++i) {
|
|
int zpi = z_ptr[i];
|
|
|
|
if (c < zpi) {
|
|
c = zpi;
|
|
}
|
|
}
|
|
|
|
c = -c;
|
|
|
|
Ref<MLPPVector> n = alg.scalar_addnv(c, z);
|
|
|
|
return softmax_normv(n);
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::adj_softmax_normm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> n = z->duplicate_fast();
|
|
|
|
Size2i size = z->size();
|
|
|
|
Ref<MLPPVector> row_rmp;
|
|
row_rmp.instance();
|
|
row_rmp->resize(size.x);
|
|
|
|
for (int i = 0; i < size.y; ++i) {
|
|
z->row_get_into_mlpp_vector(i, row_rmp);
|
|
|
|
Ref<MLPPVector> nv = adj_softmax_normv(row_rmp);
|
|
|
|
n->row_set_mlpp_vector(i, nv);
|
|
}
|
|
|
|
return n;
|
|
}
|
|
|
|
real_t MLPPActivation::adj_softmax_derivr(real_t z) {
|
|
return 0;
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::adj_softmax_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
int size = z->size();
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t c = -Math_INF;
|
|
|
|
for (int i = 0; i < size; ++i) {
|
|
int zpi = z_ptr[i];
|
|
|
|
if (c < zpi) {
|
|
c = zpi;
|
|
}
|
|
}
|
|
|
|
c = -c;
|
|
|
|
Ref<MLPPVector> n = alg.scalar_addnv(c, z);
|
|
|
|
return adj_softmax_normv(n);
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::adj_softmax_derivm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> n = z->duplicate_fast();
|
|
|
|
Size2i size = z->size();
|
|
|
|
Ref<MLPPVector> row_rmp;
|
|
row_rmp.instance();
|
|
row_rmp->resize(size.x);
|
|
|
|
for (int i = 0; i < size.y; ++i) {
|
|
z->row_get_into_mlpp_vector(i, row_rmp);
|
|
|
|
Ref<MLPPVector> nv = adj_softmax_derivv(row_rmp);
|
|
|
|
n->row_set_mlpp_vector(i, nv);
|
|
}
|
|
|
|
return n;
|
|
}
|
|
|
|
//SOFTMAX DERIV
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::softmax_deriv_normv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a = softmax_normv(z);
|
|
|
|
int z_size = z->size();
|
|
int a_size = a->size();
|
|
|
|
Ref<MLPPMatrix> deriv;
|
|
deriv.instance();
|
|
deriv->resize(Size2i(a_size, a_size));
|
|
|
|
const real_t *a_ptr = a->ptr();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
for (int j = 0; j < z_size; ++j) {
|
|
if (i == j) {
|
|
deriv->element_set(i, j, a_ptr[i] * (1 - a_ptr[i]));
|
|
} else {
|
|
deriv->element_set(i, j, -a_ptr[i] * a_ptr[j]);
|
|
}
|
|
}
|
|
}
|
|
|
|
return deriv;
|
|
}
|
|
Vector<Ref<MLPPMatrix>> MLPPActivation::softmax_deriv_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
int z_size_y = z->size().y;
|
|
|
|
Ref<MLPPMatrix> a = softmax_normm(z);
|
|
int a_size_y = a->size().y;
|
|
int a_size_x = a->size().x;
|
|
|
|
Vector<Ref<MLPPMatrix>> deriv;
|
|
deriv.resize(a_size_y);
|
|
|
|
Ref<MLPPVector> a_i_tmp;
|
|
a_i_tmp.instance();
|
|
a_i_tmp->resize(a_size_x);
|
|
|
|
Ref<MLPPVector> a_j_tmp;
|
|
a_j_tmp.instance();
|
|
a_j_tmp->resize(a_size_x);
|
|
|
|
for (int i = 0; i < deriv.size(); ++i) {
|
|
Ref<MLPPMatrix> d;
|
|
d.instance();
|
|
d->resize(Size2i(a_size_x, z_size_y));
|
|
|
|
for (int j = 0; j < z_size_y; ++j) {
|
|
a->row_get_into_mlpp_vector(i, a_i_tmp);
|
|
|
|
if (i == j) {
|
|
Ref<MLPPVector> d_j = alg.subtractionnv(a_i_tmp, alg.hadamard_productnv(a_i_tmp, a_i_tmp));
|
|
d->row_set_mlpp_vector(j, d_j);
|
|
} else {
|
|
a->row_get_into_mlpp_vector(j, a_j_tmp);
|
|
Ref<MLPPVector> d_j = alg.scalar_multiplynv(-1, alg.hadamard_productnv(a_i_tmp, a_j_tmp));
|
|
d->row_set_mlpp_vector(j, d_j);
|
|
}
|
|
}
|
|
|
|
deriv.write[i] = d;
|
|
}
|
|
|
|
return deriv;
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::softmax_deriv_derivv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a = softmax_normv(z);
|
|
|
|
int z_size = z->size();
|
|
int a_size = a->size();
|
|
|
|
Ref<MLPPMatrix> deriv;
|
|
deriv.instance();
|
|
deriv->resize(Size2i(a_size, a_size));
|
|
|
|
const real_t *a_ptr = a->ptr();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
for (int j = 0; j < z_size; ++j) {
|
|
if (i == j) {
|
|
deriv->element_set(i, j, a_ptr[i] * (1 - a_ptr[i]));
|
|
} else {
|
|
deriv->element_set(i, j, -a_ptr[i] * a_ptr[j]);
|
|
}
|
|
}
|
|
}
|
|
|
|
return deriv;
|
|
}
|
|
Vector<Ref<MLPPMatrix>> MLPPActivation::softmax_deriv_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
int z_size_y = z->size().y;
|
|
|
|
Ref<MLPPMatrix> a = softmax_normm(z);
|
|
int a_size_y = a->size().y;
|
|
int a_size_x = a->size().x;
|
|
|
|
Vector<Ref<MLPPMatrix>> deriv;
|
|
deriv.resize(a_size_y);
|
|
|
|
Ref<MLPPVector> a_i_tmp;
|
|
a_i_tmp.instance();
|
|
a_i_tmp->resize(a_size_x);
|
|
|
|
Ref<MLPPVector> a_j_tmp;
|
|
a_j_tmp.instance();
|
|
a_j_tmp->resize(a_size_x);
|
|
|
|
for (int i = 0; i < deriv.size(); ++i) {
|
|
Ref<MLPPMatrix> d;
|
|
d.instance();
|
|
d->resize(Size2i(a_size_x, z_size_y));
|
|
|
|
for (int j = 0; j < z_size_y; ++j) {
|
|
a->row_get_into_mlpp_vector(i, a_i_tmp);
|
|
|
|
if (i == j) {
|
|
Ref<MLPPVector> d_j = alg.subtractionnv(a_i_tmp, alg.hadamard_productnv(a_i_tmp, a_i_tmp));
|
|
d->row_set_mlpp_vector(j, d_j);
|
|
} else {
|
|
a->row_get_into_mlpp_vector(j, a_j_tmp);
|
|
Ref<MLPPVector> d_j = alg.scalar_multiplynv(-1, alg.hadamard_productnv(a_i_tmp, a_j_tmp));
|
|
d->row_set_mlpp_vector(j, d_j);
|
|
}
|
|
}
|
|
|
|
deriv.write[i] = d;
|
|
}
|
|
|
|
return deriv;
|
|
}
|
|
|
|
//SOFTPLUS
|
|
|
|
real_t MLPPActivation::softplus_normr(real_t z) {
|
|
return Math::log(1 + exp(z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::softplus_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognv(alg.additionnv(alg.onevecnv(z->size()), alg.expnv(z)));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::softplus_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognv(alg.additionnv(alg.onematnm(z->size().y, z->size().x), alg.expnv(z)));
|
|
}
|
|
|
|
real_t MLPPActivation::softplus_derivr(real_t z) {
|
|
return sigmoid_normr(z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::softplus_derivv(const Ref<MLPPVector> &z) {
|
|
return sigmoid_normv(z);
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::softplus_derivm(const Ref<MLPPMatrix> &z) {
|
|
return sigmoid_normm(z);
|
|
}
|
|
|
|
//SOFTSIGN
|
|
|
|
real_t MLPPActivation::softsign_normr(real_t z) {
|
|
return z / (1 + abs(z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::softsign_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(z, alg.additionnv(alg.onevecnv(z->size()), alg.absv(z)));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::softsign_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(z, alg.additionnv(alg.onematnm(z->size().y, z->size().x), alg.absnm(z)));
|
|
}
|
|
|
|
real_t MLPPActivation::softsign_derivr(real_t z) {
|
|
return 1 / ((1 + abs(z)) * (1 + abs(z)));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::softsign_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.onevecnv(z->size()), alg.exponentiatenv(alg.additionnv(alg.onevecnv(z->size()), alg.absv(z)), 2));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::softsign_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), alg.exponentiatenv(alg.additionnm(alg.onematnm(z->size().y, z->size().x), alg.absnm(z)), 2));
|
|
}
|
|
|
|
//GAUSSIANCDF
|
|
|
|
real_t MLPPActivation::gaussian_cdf_normr(real_t z) {
|
|
return 0.5 * (1 + erf(z / sqrt(2)));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::gaussian_cdf_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynv(0.5, alg.additionnv(alg.onevecnv(z->size()), alg.erfnv(alg.scalar_multiplynv(1 / sqrt(2), z))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::gaussian_cdf_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynm(0.5, alg.additionnm(alg.onematnm(z->size().y, z->size().x), alg.erfnm(alg.scalar_multiplynm(1 / sqrt(2), z))));
|
|
}
|
|
|
|
real_t MLPPActivation::gaussian_cdf_derivr(real_t z) {
|
|
return (1 / sqrt(2 * Math_PI)) * exp(-z * z / 2);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::gaussian_cdf_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynv(1 / Math::sqrt(2 * Math_PI), alg.expnv(alg.scalar_multiplynv(-1 / 2.0, alg.hadamard_productnv(z, z))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::gaussian_cdf_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynm(1 / Math::sqrt(2 * Math_PI), alg.expnm(alg.scalar_multiplynm(-1 / 2.0, alg.hadamard_productnm(z, z))));
|
|
}
|
|
|
|
//CLOGLOG
|
|
|
|
real_t MLPPActivation::cloglog_normr(real_t z) {
|
|
return 1 - exp(-exp(z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::cloglog_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynv(-1, alg.scalar_addnv(-1, alg.expnv(alg.scalar_multiplynv(-1, alg.expnv(z)))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::cloglog_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynm(-1, alg.scalar_addnm(-1, alg.expnm(alg.scalar_multiplynm(-1, alg.expnm(z)))));
|
|
}
|
|
|
|
real_t MLPPActivation::cloglog_derivr(real_t z) {
|
|
return exp(z - exp(z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::cloglog_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.expnv(alg.scalar_multiplynv(-1, alg.expnv(z)));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::cloglog_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.expnm(alg.scalar_multiplynm(-1, alg.expnm(z)));
|
|
}
|
|
|
|
//LOGIT
|
|
|
|
real_t MLPPActivation::logit_normr(real_t z) {
|
|
return Math::log(z / (1 - z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::logit_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognv(alg.division_element_wisenv(z, alg.subtractionnv(alg.onevecnv(z->size()), z)));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::logit_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognm(alg.division_element_wisenvnm(z, alg.subtractionnm(alg.onematnm(z->size().y, z->size().x), z)));
|
|
}
|
|
|
|
real_t MLPPActivation::logit_derivr(real_t z) {
|
|
return 1 / z - 1 / (z - 1);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::logit_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.subtractionnv(
|
|
alg.division_element_wisenv(alg.onevecnv(z->size()), z),
|
|
alg.division_element_wisenv(alg.onevecnv(z->size()), alg.subtractionnv(z, alg.onevecnv(z->size()))));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::logit_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.subtractionnm(
|
|
alg.division_element_wisenvnm(
|
|
alg.onematnm(z->size().y, z->size().x), z),
|
|
alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x),
|
|
alg.subtractionnm(z, alg.onematnm(z->size().y, z->size().x))));
|
|
}
|
|
|
|
//UNITSTEP
|
|
|
|
real_t MLPPActivation::unit_step_normr(real_t z) {
|
|
return z < 0 ? 0 : 1;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::unit_step_normv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = unit_step_normr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::unit_step_normm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = unit_step_normr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
real_t MLPPActivation::unit_step_derivr(real_t z) {
|
|
return 0;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::unit_step_derivv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
a->fill(0);
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::unit_step_derivm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
a->fill(0);
|
|
|
|
return a;
|
|
}
|
|
|
|
//SWISH
|
|
|
|
real_t MLPPActivation::swish_normr(real_t z) {
|
|
return z * sigmoid_normr(z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::swish_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnv(z, sigmoid_normv(z));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::swish_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnv(z, sigmoid_normm(z));
|
|
}
|
|
|
|
real_t MLPPActivation::swish_derivr(real_t z) {
|
|
return swish_normr(z) + sigmoid_normr(z) * (1 - swish_normr(z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::swish_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.additionnv(swish_normv(z), alg.subtractionnv(sigmoid_normv(z), alg.hadamard_productnv(sigmoid_normv(z), swish_normv(z))));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::swish_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.additionnv(swish_normm(z), alg.subtractionnv(sigmoid_normm(z), alg.hadamard_productnm(sigmoid_normm(z), swish_normm(z))));
|
|
}
|
|
|
|
//MISH
|
|
|
|
real_t MLPPActivation::mish_normr(real_t z) {
|
|
return z * tanh(softplus_normr(z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::mish_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnv(z, tanh_normv(softplus_normv(z)));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::mish_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnm(z, tanh_normm(softplus_normm(z)));
|
|
}
|
|
|
|
real_t MLPPActivation::mish_derivr(real_t z) {
|
|
return sech_normr(softplus_normr(z)) * sech_normr(softplus_normr(z)) * z * sigmoid_normr(z) + mish_normr(z) / z;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::mish_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.additionnv(
|
|
alg.hadamard_productnv(
|
|
alg.hadamard_productnv(
|
|
alg.hadamard_productnv(
|
|
sech_normv(softplus_normv(z)), sech_normv(softplus_normv(z))),
|
|
z),
|
|
sigmoid_normv(z)),
|
|
alg.division_element_wisenv(mish_normv(z), z));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::mish_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.additionnv(
|
|
alg.hadamard_productnm(
|
|
alg.hadamard_productnm(
|
|
alg.hadamard_productnm(
|
|
sech_normm(softplus_normm(z)), sech_normm(softplus_normm(z))),
|
|
z),
|
|
sigmoid_normm(z)),
|
|
alg.division_element_wisenvnm(mish_normm(z), z));
|
|
}
|
|
|
|
//SINC
|
|
|
|
real_t MLPPActivation::sinc_normr(real_t z) {
|
|
return Math::sin(z) / z;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::sinc_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.sinnv(z), z);
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sinc_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.sinnm(z), z);
|
|
}
|
|
|
|
real_t MLPPActivation::sinc_derivr(real_t z) {
|
|
return (z * Math::cos(z) - Math::sin(z)) / (z * z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::sinc_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.subtractionnv(alg.hadamard_productnv(z, alg.cosnv(z)), alg.sinnv(z)), alg.hadamard_productnv(z, z));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sinc_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.subtractionnm(alg.hadamard_productnm(z, alg.cosnm(z)), alg.sinnm(z)), alg.hadamard_productnm(z, z));
|
|
}
|
|
|
|
//RELU
|
|
|
|
real_t MLPPActivation::relu_normr(real_t z) {
|
|
return fmax(0, z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::relu_normv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = relu_normr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::relu_normm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = relu_normr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
real_t MLPPActivation::relu_derivr(real_t z) {
|
|
if (z <= 0) {
|
|
return 0;
|
|
} else {
|
|
return 1;
|
|
}
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::relu_derivv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = relu_derivr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::relu_derivm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = relu_derivr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
//LEAKYRELU
|
|
|
|
real_t MLPPActivation::leaky_relu_normr(real_t z, real_t c) {
|
|
return fmax(c * z, z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::leaky_relu_normv(const Ref<MLPPVector> &z, real_t c) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = leaky_relu_normr(z_ptr[i], c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::leaky_relu_normm(const Ref<MLPPMatrix> &z, real_t c) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = leaky_relu_normr(z_ptr[i], c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
real_t MLPPActivation::leaky_relu_derivr(real_t z, real_t c) {
|
|
if (z <= 0) {
|
|
return c;
|
|
} else {
|
|
return 1;
|
|
}
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::leaky_relu_derivv(const Ref<MLPPVector> &z, real_t c) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = leaky_relu_derivr(z_ptr[i], c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::leaky_relu_derivm(const Ref<MLPPMatrix> &z, real_t c) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = leaky_relu_derivr(z_ptr[i], c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
//ELU
|
|
|
|
real_t MLPPActivation::elu_normr(real_t z, real_t c) {
|
|
if (z >= 0) {
|
|
return z;
|
|
} else {
|
|
return c * (exp(z) - 1);
|
|
}
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::elu_normv(const Ref<MLPPVector> &z, real_t c) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = elu_normr(z_ptr[i], c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::elu_normm(const Ref<MLPPMatrix> &z, real_t c) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = elu_normr(z_ptr[i], c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
real_t MLPPActivation::elu_derivr(real_t z, real_t c) {
|
|
if (z <= 0) {
|
|
return c * exp(z);
|
|
} else {
|
|
return 1;
|
|
}
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::elu_derivv(const Ref<MLPPVector> &z, real_t c) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = elu_derivr(z_ptr[i], c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::elu_derivm(const Ref<MLPPMatrix> &z, real_t c) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = elu_derivr(z_ptr[i], c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
//SELU
|
|
|
|
real_t MLPPActivation::selu_normr(real_t z, real_t lambda, real_t c) {
|
|
return lambda * elu_normr(z, c);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::selu_normv(const Ref<MLPPVector> &z, real_t lambda, real_t c) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = selu_normr(z_ptr[i], lambda, c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::selu_normm(const Ref<MLPPMatrix> &z, real_t lambda, real_t c) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = selu_normr(z_ptr[i], lambda, c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
real_t MLPPActivation::selu_derivr(real_t z, real_t lambda, real_t c) {
|
|
return elu_derivr(z, c);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::selu_derivv(const Ref<MLPPVector> &z, real_t lambda, real_t c) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = selu_derivr(z_ptr[i], lambda, c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::selu_derivm(const Ref<MLPPMatrix> &z, real_t lambda, real_t c) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = selu_derivr(z_ptr[i], lambda, c);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
//GELU
|
|
|
|
real_t MLPPActivation::gelu_normr(real_t z) {
|
|
return 0.5 * z * (1 + tanh(sqrt(2 / Math_PI) * (z + 0.044715 * Math::pow(z, 3))));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::gelu_normv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = gelu_normr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::gelu_normm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = gelu_normr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
real_t MLPPActivation::gelu_derivr(real_t z) {
|
|
return 0.5 * tanh_normr(0.0356774 * Math::pow(z, 3) + 0.797885 * z) + (0.0535161 * Math::pow(z, 3) + 0.398942 * z) * Math::pow(sech_normr(0.0356774 * Math::pow(z, 3) + 0.797885 * z), 2) + 0.5;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::gelu_derivv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = gelu_derivr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::gelu_derivm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = gelu_derivr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
//SIGN
|
|
|
|
real_t MLPPActivation::sign_normr(real_t z) {
|
|
if (z < 0) {
|
|
return -1;
|
|
} else if (z == 0) {
|
|
return 0;
|
|
} else {
|
|
return 1;
|
|
}
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::sign_normv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = sign_normr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sign_normm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = sign_normr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
real_t MLPPActivation::sign_derivr(real_t z) {
|
|
return 0;
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::sign_derivv(const Ref<MLPPVector> &z) {
|
|
Ref<MLPPVector> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_size = z->size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_size; ++i) {
|
|
a_ptr[i] = sign_derivr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sign_derivm(const Ref<MLPPMatrix> &z) {
|
|
Ref<MLPPMatrix> a;
|
|
a.instance();
|
|
a->resize(z->size());
|
|
|
|
int z_data_size = z->data_size();
|
|
|
|
const real_t *z_ptr = z->ptr();
|
|
real_t *a_ptr = a->ptrw();
|
|
|
|
for (int i = 0; i < z_data_size; ++i) {
|
|
a_ptr[i] = sign_derivr(z_ptr[i]);
|
|
}
|
|
|
|
return a;
|
|
}
|
|
|
|
//SINH
|
|
|
|
real_t MLPPActivation::sinh_normr(real_t z) {
|
|
return 0.5 * (Math::exp(z) - Math::exp(-z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::sinh_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
return alg.scalar_multiplynv(0.5, alg.subtractionnv(alg.expnv(z), alg.expnv(alg.scalar_multiplynv(-1, z))));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sinh_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
return alg.scalar_multiplynm(0.5, alg.subtractionnm(alg.expnm(z), alg.expnm(alg.scalar_multiplynm(-1, z))));
|
|
}
|
|
|
|
real_t MLPPActivation::sinh_derivr(real_t z) {
|
|
return cosh_normr(z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::sinh_derivv(const Ref<MLPPVector> &z) {
|
|
return cosh_normv(z);
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sinh_derivm(const Ref<MLPPMatrix> &z) {
|
|
return cosh_normm(z);
|
|
}
|
|
|
|
//COSH
|
|
|
|
real_t MLPPActivation::cosh_normr(real_t z) {
|
|
return 0.5 * (Math::exp(z) + Math::exp(-z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::cosh_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
return alg.scalar_multiplynv(0.5, alg.additionnv(alg.expnv(z), alg.expnv(alg.scalar_multiplynv(-1, z))));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::cosh_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
return alg.scalar_multiplynm(0.5, alg.additionnm(alg.expnm(z), alg.expnm(alg.scalar_multiplynm(-1, z))));
|
|
}
|
|
|
|
real_t MLPPActivation::cosh_derivr(real_t z) {
|
|
return sinh_normr(z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::cosh_derivv(const Ref<MLPPVector> &z) {
|
|
return sinh_normv(z);
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::cosh_derivm(const Ref<MLPPMatrix> &z) {
|
|
return sinh_normm(z);
|
|
}
|
|
|
|
//TANH
|
|
|
|
real_t MLPPActivation::tanh_normr(real_t z) {
|
|
return (Math::exp(z) - Math::exp(-z)) / (Math::exp(z) + Math::exp(-z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::tanh_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.subtractionnv(alg.expnv(z), alg.expnv(alg.scalar_multiplynv(-1, z))), alg.additionnv(alg.expnv(z), alg.expnv(alg.scalar_multiplynv(-1, z))));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::tanh_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.subtractionnm(alg.expnm(z), alg.expnm(alg.scalar_multiplynm(-1, z))), alg.additionnm(alg.expnm(z), alg.expnm(alg.scalar_multiplynm(-1, z))));
|
|
}
|
|
|
|
real_t MLPPActivation::tanh_derivr(real_t z) {
|
|
return 1 - tanh(z) * tanh(z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::tanh_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynv(-1, alg.scalar_addnv(-1, alg.hadamard_productnv(tanh_normv(z), tanh_normv(z))));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::tanh_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynm(-1, alg.scalar_addnm(-1, alg.hadamard_productnm(tanh_normm(z), tanh_normm(z))));
|
|
}
|
|
|
|
//CSCH
|
|
|
|
real_t MLPPActivation::csch_normr(real_t z) {
|
|
return 1 / sinh(z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::csch_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.onevecnv(z->size()), sinh_normv(z));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::csch_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), sinh_normm(z));
|
|
}
|
|
|
|
real_t MLPPActivation::csch_derivr(real_t z) {
|
|
return -csch_normr(z) * coth_normr(z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::csch_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnv(alg.scalar_multiplynv(-1, csch_normv(z)), coth_normv(z));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::csch_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnm(alg.scalar_multiplynm(-1, csch_normm(z)), coth_normm(z));
|
|
}
|
|
|
|
//SECH
|
|
|
|
real_t MLPPActivation::sech_normr(real_t z) {
|
|
return 1 / cosh(z);
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::sech_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.onevecnv(z->size()), cosh_normv(z));
|
|
|
|
// return activation(z, deriv, static_cast<void (*)(real_t, bool)>(&sech));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sech_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), cosh_normm(z));
|
|
|
|
// return activation(z, deriv, static_cast<void (*)(real_t, bool)>(&sech));
|
|
}
|
|
|
|
real_t MLPPActivation::sech_derivr(real_t z) {
|
|
return -sech_normr(z) * tanh_normr(z);
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::sech_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnv(alg.scalar_multiplynv(-1, sech_normv(z)), tanh_normv(z));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::sech_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnm(alg.scalar_multiplynm(-1, sech_normm(z)), tanh_normm(z));
|
|
}
|
|
|
|
//COTH
|
|
|
|
real_t MLPPActivation::coth_normr(real_t z) {
|
|
return 1 / tanh(z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::coth_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.onevecnv(z->size()), tanh_normv(z));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::coth_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), tanh_normm(z));
|
|
}
|
|
|
|
real_t MLPPActivation::coth_derivr(real_t z) {
|
|
return -csch_normr(z) * csch_normr(z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::coth_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnv(alg.scalar_multiplynv(-1, csch_normv(z)), csch_normv(z));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::coth_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.hadamard_productnm(alg.scalar_multiplynm(-1, csch_normm(z)), csch_normm(z));
|
|
}
|
|
|
|
//ARSINH
|
|
|
|
real_t MLPPActivation::arsinh_normr(real_t z) {
|
|
return Math::log(z + sqrt(z * z + 1));
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::arsinh_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognv(alg.additionnv(z, alg.sqrtnv(alg.additionnv(alg.hadamard_productnv(z, z), alg.onevecnv(z->size())))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::arsinh_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognm(alg.additionnm(z, alg.sqrtnm(alg.additionnm(alg.hadamard_productnm(z, z), alg.onematnm(z->size().y, z->size().x)))));
|
|
}
|
|
|
|
real_t MLPPActivation::arsinh_derivr(real_t z) {
|
|
return 1 / sqrt(z * z + 1);
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::arsinh_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.onevecnv(z->size()), alg.sqrtnv(alg.additionnv(alg.hadamard_productnv(z, z), alg.onevecnv(z->size()))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::arsinh_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), alg.sqrtnm(alg.additionnm(alg.hadamard_productnm(z, z), alg.onematnm(z->size().y, z->size().x))));
|
|
}
|
|
|
|
//ARCOSH
|
|
|
|
real_t MLPPActivation::arcosh_normr(real_t z) {
|
|
return Math::log(z + sqrt(z * z - 1));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::arcosh_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognv(alg.additionnv(z, alg.sqrtnv(alg.subtractionnv(alg.hadamard_productnv(z, z), alg.onevecnv(z->size())))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::arcosh_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognm(alg.additionnm(z, alg.sqrtnm(alg.subtractionnm(alg.hadamard_productnm(z, z), alg.onematnm(z->size().y, z->size().x)))));
|
|
}
|
|
|
|
real_t MLPPActivation::arcosh_derivr(real_t z) {
|
|
return 1 / sqrt(z * z - 1);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::arcosh_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.onevecnv(z->size()), alg.sqrtnv(alg.subtractionnv(alg.hadamard_productnv(z, z), alg.onevecnv(z->size()))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::arcosh_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), alg.sqrtnm(alg.subtractionnm(alg.hadamard_productnm(z, z), alg.onematnm(z->size().y, z->size().x))));
|
|
}
|
|
|
|
//ARTANH
|
|
|
|
real_t MLPPActivation::artanh_normr(real_t z) {
|
|
return 0.5 * Math::log((1 + z) / (1 - z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::artanh_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynv(0.5, alg.lognv(alg.division_element_wisenv(alg.additionnv(alg.onevecnv(z->size()), z), alg.subtractionnv(alg.onevecnv(z->size()), z))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::artanh_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynm(0.5, alg.lognm(alg.division_element_wisenvnm(alg.additionnm(alg.onematnm(z->size().y, z->size().x), z), alg.subtractionnm(alg.onematnm(z->size().y, z->size().x), z))));
|
|
}
|
|
|
|
real_t MLPPActivation::artanh_derivr(real_t z) {
|
|
return 1 / (1 - z * z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::artanh_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.onevecnv(z->size()), alg.subtractionnv(alg.onevecnv(z->size()), alg.hadamard_productnv(z, z)));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::artanh_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), alg.subtractionnv(alg.onematnm(z->size().y, z->size().x), alg.hadamard_productnm(z, z)));
|
|
}
|
|
|
|
//ARCSCH
|
|
|
|
real_t MLPPActivation::arcsch_normr(real_t z) {
|
|
return Math::log(sqrt(1 + (1 / (z * z))) + (1 / z));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::arcsch_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognv(
|
|
alg.additionnv(
|
|
alg.sqrtnv(
|
|
alg.additionnv(
|
|
alg.onevecnv(z->size()),
|
|
alg.division_element_wisenv(alg.onevecnv(z->size()), alg.hadamard_productnv(z, z)))),
|
|
alg.division_element_wisenv(alg.onevecnv(z->size()), z)));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::arcsch_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognm(
|
|
alg.additionnm(
|
|
alg.sqrtnm(
|
|
alg.additionnm(alg.onematnm(z->size().y, z->size().x),
|
|
alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), alg.hadamard_productnm(z, z)))),
|
|
alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), z)));
|
|
}
|
|
|
|
real_t MLPPActivation::arcsch_derivr(real_t z) {
|
|
return -1 / ((z * z) * sqrt(1 + (1 / (z * z))));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::arcsch_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(
|
|
alg.fullnv(z->size(), -1),
|
|
alg.hadamard_productnm(
|
|
alg.hadamard_productnv(z, z),
|
|
alg.sqrtnv(alg.additionnv(alg.onevecnv(z->size()), alg.division_element_wisenv(alg.onevecnv(z->size()), alg.hadamard_productnv(z, z))))));
|
|
}
|
|
Ref<MLPPMatrix> MLPPActivation::arcsch_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(
|
|
alg.fullnm(z->size().y, z->size().x, -1),
|
|
alg.hadamard_productnm(alg.hadamard_productnm(z, z),
|
|
alg.sqrtnm(alg.additionnm(alg.onematnm(z->size().y, z->size().x),
|
|
alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), alg.hadamard_productnm(z, z))))));
|
|
}
|
|
|
|
//ARSECH
|
|
|
|
real_t MLPPActivation::arsech_normr(real_t z) {
|
|
return Math::log((1 / z) + ((1 / z) + 1) * ((1 / z) - 1));
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::arsech_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognv(
|
|
alg.additionnv(
|
|
alg.division_element_wisenv(
|
|
alg.onevecnv(z->size()), z),
|
|
alg.hadamard_productnv(
|
|
alg.additionnv(alg.division_element_wisenv(alg.onevecnv(z->size()), z), alg.onevecnv(z->size())),
|
|
alg.subtractionnv(alg.division_element_wisenv(alg.onevecnv(z->size()), z), alg.onevecnv(z->size())))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::arsech_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.lognm(
|
|
alg.additionnm(
|
|
alg.division_element_wisenvnm(
|
|
alg.onematnm(z->size().y, z->size().x), z),
|
|
alg.hadamard_productnm(
|
|
alg.additionnm(
|
|
alg.division_element_wisenvnm(
|
|
alg.onematnm(z->size().y, z->size().x), z),
|
|
alg.onematnm(z->size().y, z->size().x)),
|
|
alg.subtractionnm(
|
|
alg.division_element_wisenvnm(
|
|
alg.onematnm(z->size().y, z->size().x), z),
|
|
alg.onematnm(z->size().y, z->size().x)))));
|
|
}
|
|
|
|
real_t MLPPActivation::arsech_derivr(real_t z) {
|
|
return -1 / (z * sqrt(1 - z * z));
|
|
}
|
|
|
|
Ref<MLPPVector> MLPPActivation::arsech_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(
|
|
alg.fullnv(z->size(), -1),
|
|
alg.hadamard_productnv(
|
|
z,
|
|
alg.sqrtnv(
|
|
alg.subtractionnv(alg.onevecnv(z->size()), alg.hadamard_productnv(z, z)))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::arsech_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(
|
|
alg.fullnm(z->size().y, z->size().x, -1),
|
|
alg.hadamard_productnm(
|
|
z,
|
|
alg.sqrtnm(alg.subtractionnm(alg.onematnm(z->size().y, z->size().x), alg.hadamard_productnm(z, z)))));
|
|
}
|
|
|
|
//ARCOTH
|
|
|
|
real_t MLPPActivation::arcoth_normr(real_t z) {
|
|
return 0.5 * Math::log((1 + z) / (z - 1));
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::arcoth_normv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynv(
|
|
0.5,
|
|
alg.lognv(alg.division_element_wisenv(alg.additionnv(alg.onevecnv(z->size()), z), alg.subtractionnv(z, alg.onevecnv(z->size())))));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::arcoth_normm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.scalar_multiplynm(
|
|
0.5,
|
|
alg.lognm(alg.division_element_wisenvnm(alg.additionnm(alg.onematnm(z->size().y, z->size().x), z), alg.subtractionnm(z, alg.onematnm(z->size().y, z->size().x)))));
|
|
}
|
|
|
|
real_t MLPPActivation::arcoth_derivr(real_t z) {
|
|
return 1 / (1 - z * z);
|
|
}
|
|
Ref<MLPPVector> MLPPActivation::arcoth_derivv(const Ref<MLPPVector> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenv(alg.onevecnv(z->size()), alg.subtractionnv(alg.onevecnv(z->size()), alg.hadamard_productnv(z, z)));
|
|
}
|
|
|
|
Ref<MLPPMatrix> MLPPActivation::arcoth_derivm(const Ref<MLPPMatrix> &z) {
|
|
MLPPLinAlg alg;
|
|
|
|
return alg.division_element_wisenvnm(alg.onematnm(z->size().y, z->size().x), alg.subtractionnm(alg.onematnm(z->size().y, z->size().x), alg.hadamard_productnm(z, z)));
|
|
}
|
|
|
|
void MLPPActivation::_bind_methods() {
|
|
ClassDB::bind_method(D_METHOD("run_activation_real", "func", "z", "deriv"), &MLPPActivation::run_activation_real, false);
|
|
ClassDB::bind_method(D_METHOD("run_activation_vector", "func", "z", "deriv"), &MLPPActivation::run_activation_vector, false);
|
|
ClassDB::bind_method(D_METHOD("run_activation_matrix", "func", "z", "deriv"), &MLPPActivation::run_activation_matrix, false);
|
|
|
|
ClassDB::bind_method(D_METHOD("run_activation_norm_real", "func", "z"), &MLPPActivation::run_activation_norm_real);
|
|
ClassDB::bind_method(D_METHOD("run_activation_norm_vector", "func", "z"), &MLPPActivation::run_activation_norm_vector);
|
|
ClassDB::bind_method(D_METHOD("run_activation_norm_matrix", "func", "z"), &MLPPActivation::run_activation_norm_matrix);
|
|
|
|
real_t run_activation_norm_real(const ActivationFunction func, const real_t z);
|
|
Ref<MLPPVector> run_activation_norm_vector(const ActivationFunction func, const Ref<MLPPVector> &z);
|
|
Ref<MLPPMatrix> run_activation_norm_matrix(const ActivationFunction func, const Ref<MLPPMatrix> &z);
|
|
|
|
ClassDB::bind_method(D_METHOD("run_activation_deriv_real", "func", "z"), &MLPPActivation::run_activation_deriv_real);
|
|
ClassDB::bind_method(D_METHOD("run_activation_deriv_vector", "func", "z"), &MLPPActivation::run_activation_deriv_vector);
|
|
ClassDB::bind_method(D_METHOD("run_activation_deriv_matrix", "func", "z"), &MLPPActivation::run_activation_deriv_matrix);
|
|
|
|
//LINEAR
|
|
|
|
ClassDB::bind_method(D_METHOD("linear_normr", "z"), &MLPPActivation::linear_normr);
|
|
ClassDB::bind_method(D_METHOD("linear_normv", "z"), &MLPPActivation::linear_normv);
|
|
ClassDB::bind_method(D_METHOD("linear_normm", "z"), &MLPPActivation::linear_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("linear_derivr", "z"), &MLPPActivation::linear_derivr);
|
|
ClassDB::bind_method(D_METHOD("linear_derivv", "z"), &MLPPActivation::linear_derivv);
|
|
ClassDB::bind_method(D_METHOD("linear_derivm", "z"), &MLPPActivation::linear_derivm);
|
|
|
|
//SIGMOID
|
|
|
|
ClassDB::bind_method(D_METHOD("sigmoid_normr", "z"), &MLPPActivation::sigmoid_normr);
|
|
ClassDB::bind_method(D_METHOD("sigmoid_normv", "z"), &MLPPActivation::sigmoid_normv);
|
|
ClassDB::bind_method(D_METHOD("sigmoid_normm", "z"), &MLPPActivation::sigmoid_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("sigmoid_derivr", "z"), &MLPPActivation::sigmoid_derivr);
|
|
ClassDB::bind_method(D_METHOD("sigmoid_derivv", "z"), &MLPPActivation::sigmoid_derivv);
|
|
ClassDB::bind_method(D_METHOD("sigmoid_derivm", "z"), &MLPPActivation::sigmoid_derivm);
|
|
|
|
//SOFTMAX
|
|
|
|
ClassDB::bind_method(D_METHOD("softmax_normr", "z"), &MLPPActivation::softmax_normr);
|
|
ClassDB::bind_method(D_METHOD("softmax_normv", "z"), &MLPPActivation::softmax_normv);
|
|
ClassDB::bind_method(D_METHOD("softmax_normm", "z"), &MLPPActivation::softmax_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("softmax_derivr", "z"), &MLPPActivation::softmax_derivr);
|
|
ClassDB::bind_method(D_METHOD("softmax_derivv", "z"), &MLPPActivation::softmax_derivv);
|
|
ClassDB::bind_method(D_METHOD("softmax_derivm", "z"), &MLPPActivation::softmax_derivm);
|
|
|
|
//ADJ_SOFTMAX
|
|
|
|
real_t adj_softmax_normr(real_t z);
|
|
Ref<MLPPVector> adj_softmax_normv(const Ref<MLPPVector> &z);
|
|
Ref<MLPPMatrix> adj_softmax_normm(const Ref<MLPPMatrix> &z);
|
|
|
|
real_t adj_softmax_derivr(real_t z);
|
|
Ref<MLPPVector> adj_softmax_derivv(const Ref<MLPPVector> &z);
|
|
Ref<MLPPMatrix> adj_softmax_derivm(const Ref<MLPPMatrix> &z);
|
|
|
|
//SOFTPLUS
|
|
|
|
ClassDB::bind_method(D_METHOD("softplus_normr", "z"), &MLPPActivation::softplus_normr);
|
|
ClassDB::bind_method(D_METHOD("softplus_normv", "z"), &MLPPActivation::softplus_normv);
|
|
ClassDB::bind_method(D_METHOD("softplus_normm", "z"), &MLPPActivation::softplus_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("softplus_derivr", "z"), &MLPPActivation::softplus_derivr);
|
|
ClassDB::bind_method(D_METHOD("softplus_derivv", "z"), &MLPPActivation::softplus_derivv);
|
|
ClassDB::bind_method(D_METHOD("softplus_derivm", "z"), &MLPPActivation::softplus_derivm);
|
|
|
|
//SOFTSIGN
|
|
|
|
ClassDB::bind_method(D_METHOD("softsign_normr", "z"), &MLPPActivation::softsign_normr);
|
|
ClassDB::bind_method(D_METHOD("softsign_normv", "z"), &MLPPActivation::softsign_normv);
|
|
ClassDB::bind_method(D_METHOD("softsign_normm", "z"), &MLPPActivation::softsign_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("softsign_derivr", "z"), &MLPPActivation::softsign_derivr);
|
|
ClassDB::bind_method(D_METHOD("softsign_derivv", "z"), &MLPPActivation::softsign_derivv);
|
|
ClassDB::bind_method(D_METHOD("softsign_derivm", "z"), &MLPPActivation::softsign_derivm);
|
|
|
|
//GAUSSIANCDF
|
|
|
|
ClassDB::bind_method(D_METHOD("gaussian_cdf_normr", "z"), &MLPPActivation::gaussian_cdf_normr);
|
|
ClassDB::bind_method(D_METHOD("gaussian_cdf_normv", "z"), &MLPPActivation::gaussian_cdf_normv);
|
|
ClassDB::bind_method(D_METHOD("gaussian_cdf_normm", "z"), &MLPPActivation::gaussian_cdf_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("gaussian_cdf_derivr", "z"), &MLPPActivation::gaussian_cdf_derivr);
|
|
ClassDB::bind_method(D_METHOD("gaussian_cdf_derivv", "z"), &MLPPActivation::gaussian_cdf_derivv);
|
|
ClassDB::bind_method(D_METHOD("gaussian_cdf_derivm", "z"), &MLPPActivation::gaussian_cdf_derivm);
|
|
|
|
//CLOGLOG
|
|
|
|
ClassDB::bind_method(D_METHOD("cloglog_normr", "z"), &MLPPActivation::cloglog_normr);
|
|
ClassDB::bind_method(D_METHOD("cloglog_normv", "z"), &MLPPActivation::cloglog_normv);
|
|
ClassDB::bind_method(D_METHOD("cloglog_normm", "z"), &MLPPActivation::cloglog_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("cloglog_derivr", "z"), &MLPPActivation::cloglog_derivr);
|
|
ClassDB::bind_method(D_METHOD("cloglog_derivv", "z"), &MLPPActivation::cloglog_derivv);
|
|
ClassDB::bind_method(D_METHOD("cloglog_derivm", "z"), &MLPPActivation::cloglog_derivm);
|
|
|
|
//LOGIT
|
|
|
|
ClassDB::bind_method(D_METHOD("logit_normr", "z"), &MLPPActivation::logit_normr);
|
|
ClassDB::bind_method(D_METHOD("logit_normv", "z"), &MLPPActivation::logit_normv);
|
|
ClassDB::bind_method(D_METHOD("logit_normm", "z"), &MLPPActivation::logit_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("logit_derivr", "z"), &MLPPActivation::logit_derivr);
|
|
ClassDB::bind_method(D_METHOD("logit_derivv", "z"), &MLPPActivation::logit_derivv);
|
|
ClassDB::bind_method(D_METHOD("logit_derivm", "z"), &MLPPActivation::logit_derivm);
|
|
|
|
//UNITSTEP
|
|
|
|
ClassDB::bind_method(D_METHOD("unit_step_normr", "z"), &MLPPActivation::unit_step_normr);
|
|
ClassDB::bind_method(D_METHOD("unit_step_normv", "z"), &MLPPActivation::unit_step_normv);
|
|
ClassDB::bind_method(D_METHOD("unit_step_normm", "z"), &MLPPActivation::unit_step_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("unit_step_derivr", "z"), &MLPPActivation::unit_step_derivr);
|
|
ClassDB::bind_method(D_METHOD("unit_step_derivv", "z"), &MLPPActivation::unit_step_derivv);
|
|
ClassDB::bind_method(D_METHOD("unit_step_derivm", "z"), &MLPPActivation::unit_step_derivm);
|
|
|
|
//SWISH
|
|
|
|
ClassDB::bind_method(D_METHOD("swish_normr", "z"), &MLPPActivation::swish_normr);
|
|
ClassDB::bind_method(D_METHOD("swish_normv", "z"), &MLPPActivation::swish_normv);
|
|
ClassDB::bind_method(D_METHOD("swish_normm", "z"), &MLPPActivation::swish_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("swish_derivr", "z"), &MLPPActivation::swish_derivr);
|
|
ClassDB::bind_method(D_METHOD("swish_derivv", "z"), &MLPPActivation::swish_derivv);
|
|
ClassDB::bind_method(D_METHOD("swish_derivm", "z"), &MLPPActivation::swish_derivm);
|
|
|
|
//MISH
|
|
|
|
ClassDB::bind_method(D_METHOD("mish_normr", "z"), &MLPPActivation::mish_normr);
|
|
ClassDB::bind_method(D_METHOD("mish_normv", "z"), &MLPPActivation::mish_normv);
|
|
ClassDB::bind_method(D_METHOD("mish_normm", "z"), &MLPPActivation::mish_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("mish_derivr", "z"), &MLPPActivation::mish_derivr);
|
|
ClassDB::bind_method(D_METHOD("mish_derivv", "z"), &MLPPActivation::mish_derivv);
|
|
ClassDB::bind_method(D_METHOD("mish_derivm", "z"), &MLPPActivation::mish_derivm);
|
|
|
|
//SINC
|
|
|
|
ClassDB::bind_method(D_METHOD("sinc_normr", "z"), &MLPPActivation::sinc_normr);
|
|
ClassDB::bind_method(D_METHOD("sinc_normv", "z"), &MLPPActivation::sinc_normv);
|
|
ClassDB::bind_method(D_METHOD("sinc_normm", "z"), &MLPPActivation::sinc_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("sinc_derivr", "z"), &MLPPActivation::sinc_derivr);
|
|
ClassDB::bind_method(D_METHOD("sinc_derivv", "z"), &MLPPActivation::sinc_derivv);
|
|
ClassDB::bind_method(D_METHOD("sinc_derivm", "z"), &MLPPActivation::sinc_derivm);
|
|
|
|
//RELU
|
|
|
|
ClassDB::bind_method(D_METHOD("relu_normr", "z"), &MLPPActivation::relu_normr);
|
|
ClassDB::bind_method(D_METHOD("relu_normv", "z"), &MLPPActivation::relu_normv);
|
|
ClassDB::bind_method(D_METHOD("relu_normm", "z"), &MLPPActivation::relu_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("relu_derivr", "z"), &MLPPActivation::relu_derivr);
|
|
ClassDB::bind_method(D_METHOD("relu_derivv", "z"), &MLPPActivation::relu_derivv);
|
|
ClassDB::bind_method(D_METHOD("relu_derivm", "z"), &MLPPActivation::relu_derivm);
|
|
|
|
//LEAKYRELU
|
|
|
|
ClassDB::bind_method(D_METHOD("leaky_relu_normr", "z"), &MLPPActivation::leaky_relu_normr);
|
|
ClassDB::bind_method(D_METHOD("leaky_relu_normv", "z"), &MLPPActivation::leaky_relu_normv);
|
|
ClassDB::bind_method(D_METHOD("leaky_relu_normm", "z"), &MLPPActivation::leaky_relu_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("leaky_relu_derivr", "z"), &MLPPActivation::leaky_relu_derivr);
|
|
ClassDB::bind_method(D_METHOD("leaky_relu_derivv", "z"), &MLPPActivation::leaky_relu_derivv);
|
|
ClassDB::bind_method(D_METHOD("leaky_relu_derivm", "z"), &MLPPActivation::leaky_relu_derivm);
|
|
|
|
//ELU
|
|
|
|
ClassDB::bind_method(D_METHOD("elu_normr", "z"), &MLPPActivation::elu_normr);
|
|
ClassDB::bind_method(D_METHOD("elu_normv", "z"), &MLPPActivation::elu_normv);
|
|
ClassDB::bind_method(D_METHOD("elu_normm", "z"), &MLPPActivation::elu_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("elu_derivr", "z"), &MLPPActivation::elu_derivr);
|
|
ClassDB::bind_method(D_METHOD("elu_derivv", "z"), &MLPPActivation::elu_derivv);
|
|
ClassDB::bind_method(D_METHOD("elu_derivm", "z"), &MLPPActivation::elu_derivm);
|
|
|
|
//SELU
|
|
|
|
ClassDB::bind_method(D_METHOD("selu_normr", "z"), &MLPPActivation::selu_normr);
|
|
ClassDB::bind_method(D_METHOD("selu_normv", "z"), &MLPPActivation::selu_normv);
|
|
ClassDB::bind_method(D_METHOD("selu_normm", "z"), &MLPPActivation::selu_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("selu_derivr", "z"), &MLPPActivation::selu_derivr);
|
|
ClassDB::bind_method(D_METHOD("selu_derivv", "z"), &MLPPActivation::selu_derivv);
|
|
ClassDB::bind_method(D_METHOD("selu_derivm", "z"), &MLPPActivation::selu_derivm);
|
|
|
|
//GELU
|
|
|
|
ClassDB::bind_method(D_METHOD("gelu_normr", "z"), &MLPPActivation::gelu_normr);
|
|
ClassDB::bind_method(D_METHOD("gelu_normv", "z"), &MLPPActivation::gelu_normv);
|
|
ClassDB::bind_method(D_METHOD("gelu_normm", "z"), &MLPPActivation::gelu_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("gelu_derivr", "z"), &MLPPActivation::gelu_derivr);
|
|
ClassDB::bind_method(D_METHOD("gelu_derivv", "z"), &MLPPActivation::gelu_derivv);
|
|
ClassDB::bind_method(D_METHOD("gelu_derivm", "z"), &MLPPActivation::gelu_derivm);
|
|
|
|
//SIGN
|
|
|
|
ClassDB::bind_method(D_METHOD("sign_normr", "z"), &MLPPActivation::sign_normr);
|
|
ClassDB::bind_method(D_METHOD("sign_normv", "z"), &MLPPActivation::sign_normv);
|
|
ClassDB::bind_method(D_METHOD("sign_normm", "z"), &MLPPActivation::sign_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("sign_derivr", "z"), &MLPPActivation::sign_derivr);
|
|
ClassDB::bind_method(D_METHOD("sign_derivv", "z"), &MLPPActivation::sign_derivv);
|
|
ClassDB::bind_method(D_METHOD("sign_derivm", "z"), &MLPPActivation::sign_derivm);
|
|
|
|
//SINH
|
|
|
|
ClassDB::bind_method(D_METHOD("sinh_normr", "z"), &MLPPActivation::sinh_normr);
|
|
ClassDB::bind_method(D_METHOD("sinh_normv", "z"), &MLPPActivation::sinh_normv);
|
|
ClassDB::bind_method(D_METHOD("sinh_normm", "z"), &MLPPActivation::sinh_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("sinh_derivr", "z"), &MLPPActivation::sinh_derivr);
|
|
ClassDB::bind_method(D_METHOD("sinh_derivv", "z"), &MLPPActivation::sinh_derivv);
|
|
ClassDB::bind_method(D_METHOD("sinh_derivm", "z"), &MLPPActivation::sinh_derivm);
|
|
|
|
//COSH
|
|
|
|
ClassDB::bind_method(D_METHOD("cosh_normr", "z"), &MLPPActivation::cosh_normr);
|
|
ClassDB::bind_method(D_METHOD("cosh_normv", "z"), &MLPPActivation::cosh_normv);
|
|
ClassDB::bind_method(D_METHOD("cosh_normm", "z"), &MLPPActivation::cosh_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("cosh_derivr", "z"), &MLPPActivation::cosh_derivr);
|
|
ClassDB::bind_method(D_METHOD("cosh_derivv", "z"), &MLPPActivation::cosh_derivv);
|
|
ClassDB::bind_method(D_METHOD("cosh_derivm", "z"), &MLPPActivation::cosh_derivm);
|
|
|
|
//TANH
|
|
|
|
ClassDB::bind_method(D_METHOD("tanh_normr", "z"), &MLPPActivation::tanh_normr);
|
|
ClassDB::bind_method(D_METHOD("tanh_normv", "z"), &MLPPActivation::tanh_normv);
|
|
ClassDB::bind_method(D_METHOD("tanh_normm", "z"), &MLPPActivation::tanh_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("tanh_derivr", "z"), &MLPPActivation::tanh_derivr);
|
|
ClassDB::bind_method(D_METHOD("tanh_derivv", "z"), &MLPPActivation::tanh_derivv);
|
|
ClassDB::bind_method(D_METHOD("tanh_derivm", "z"), &MLPPActivation::tanh_derivm);
|
|
|
|
//CSCH
|
|
|
|
ClassDB::bind_method(D_METHOD("csch_normr", "z"), &MLPPActivation::csch_normr);
|
|
ClassDB::bind_method(D_METHOD("csch_normv", "z"), &MLPPActivation::csch_normv);
|
|
ClassDB::bind_method(D_METHOD("csch_normm", "z"), &MLPPActivation::csch_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("csch_derivr", "z"), &MLPPActivation::csch_derivr);
|
|
ClassDB::bind_method(D_METHOD("csch_derivv", "z"), &MLPPActivation::csch_derivv);
|
|
ClassDB::bind_method(D_METHOD("csch_derivm", "z"), &MLPPActivation::csch_derivm);
|
|
|
|
//SECH
|
|
|
|
ClassDB::bind_method(D_METHOD("sech_normr", "z"), &MLPPActivation::sech_normr);
|
|
ClassDB::bind_method(D_METHOD("sech_normv", "z"), &MLPPActivation::sech_normv);
|
|
ClassDB::bind_method(D_METHOD("sech_normm", "z"), &MLPPActivation::sech_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("sech_derivr", "z"), &MLPPActivation::sech_derivr);
|
|
ClassDB::bind_method(D_METHOD("sech_derivv", "z"), &MLPPActivation::sech_derivv);
|
|
ClassDB::bind_method(D_METHOD("sech_derivm", "z"), &MLPPActivation::sech_derivm);
|
|
|
|
//COTH
|
|
|
|
ClassDB::bind_method(D_METHOD("coth_normr", "z"), &MLPPActivation::coth_normr);
|
|
ClassDB::bind_method(D_METHOD("coth_normv", "z"), &MLPPActivation::coth_normv);
|
|
ClassDB::bind_method(D_METHOD("coth_normm", "z"), &MLPPActivation::coth_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("coth_derivr", "z"), &MLPPActivation::coth_derivr);
|
|
ClassDB::bind_method(D_METHOD("coth_derivv", "z"), &MLPPActivation::coth_derivv);
|
|
ClassDB::bind_method(D_METHOD("coth_derivm", "z"), &MLPPActivation::coth_derivm);
|
|
|
|
//ARSINH
|
|
|
|
ClassDB::bind_method(D_METHOD("arsinh_normr", "z"), &MLPPActivation::arsinh_normr);
|
|
ClassDB::bind_method(D_METHOD("arsinh_normv", "z"), &MLPPActivation::arsinh_normv);
|
|
ClassDB::bind_method(D_METHOD("arsinh_normm", "z"), &MLPPActivation::arsinh_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("arsinh_derivr", "z"), &MLPPActivation::arsinh_derivr);
|
|
ClassDB::bind_method(D_METHOD("arsinh_derivv", "z"), &MLPPActivation::arsinh_derivv);
|
|
ClassDB::bind_method(D_METHOD("arsinh_derivm", "z"), &MLPPActivation::arsinh_derivm);
|
|
|
|
//ARCOSH
|
|
|
|
ClassDB::bind_method(D_METHOD("arcosh_normr", "z"), &MLPPActivation::arcosh_normr);
|
|
ClassDB::bind_method(D_METHOD("arcosh_normv", "z"), &MLPPActivation::arcosh_normv);
|
|
ClassDB::bind_method(D_METHOD("arcosh_normm", "z"), &MLPPActivation::arcosh_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("arcosh_derivr", "z"), &MLPPActivation::arcosh_derivr);
|
|
ClassDB::bind_method(D_METHOD("arcosh_derivv", "z"), &MLPPActivation::arcosh_derivv);
|
|
ClassDB::bind_method(D_METHOD("arcosh_derivm", "z"), &MLPPActivation::arcosh_derivm);
|
|
|
|
//ARTANH
|
|
|
|
ClassDB::bind_method(D_METHOD("artanh_normr", "z"), &MLPPActivation::artanh_normr);
|
|
ClassDB::bind_method(D_METHOD("artanh_normv", "z"), &MLPPActivation::artanh_normv);
|
|
ClassDB::bind_method(D_METHOD("artanh_normm", "z"), &MLPPActivation::artanh_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("artanh_derivr", "z"), &MLPPActivation::artanh_derivr);
|
|
ClassDB::bind_method(D_METHOD("artanh_derivv", "z"), &MLPPActivation::artanh_derivv);
|
|
ClassDB::bind_method(D_METHOD("artanh_derivm", "z"), &MLPPActivation::artanh_derivm);
|
|
|
|
//ARCSCH
|
|
|
|
ClassDB::bind_method(D_METHOD("arcsch_normr", "z"), &MLPPActivation::arcsch_normr);
|
|
ClassDB::bind_method(D_METHOD("arcsch_normv", "z"), &MLPPActivation::arcsch_normv);
|
|
ClassDB::bind_method(D_METHOD("arcsch_normm", "z"), &MLPPActivation::arcsch_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("arcsch_derivr", "z"), &MLPPActivation::arcsch_derivr);
|
|
ClassDB::bind_method(D_METHOD("arcsch_derivv", "z"), &MLPPActivation::arcsch_derivv);
|
|
ClassDB::bind_method(D_METHOD("arcsch_derivm", "z"), &MLPPActivation::arcsch_derivm);
|
|
|
|
//ARSECH
|
|
|
|
ClassDB::bind_method(D_METHOD("arsech_normr", "z"), &MLPPActivation::arsech_normr);
|
|
ClassDB::bind_method(D_METHOD("arsech_normv", "z"), &MLPPActivation::arsech_normv);
|
|
ClassDB::bind_method(D_METHOD("arsech_normm", "z"), &MLPPActivation::arsech_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("arsech_derivr", "z"), &MLPPActivation::arsech_derivr);
|
|
ClassDB::bind_method(D_METHOD("arsech_derivv", "z"), &MLPPActivation::arsech_derivv);
|
|
ClassDB::bind_method(D_METHOD("arsech_derivm", "z"), &MLPPActivation::arsech_derivm);
|
|
|
|
//ARCOTH
|
|
|
|
ClassDB::bind_method(D_METHOD("arcoth_normr", "z"), &MLPPActivation::arcoth_normr);
|
|
ClassDB::bind_method(D_METHOD("arcoth_normv", "z"), &MLPPActivation::arcoth_normv);
|
|
ClassDB::bind_method(D_METHOD("arcoth_normm", "z"), &MLPPActivation::arcoth_normm);
|
|
|
|
ClassDB::bind_method(D_METHOD("arcoth_derivr", "z"), &MLPPActivation::arcoth_derivr);
|
|
ClassDB::bind_method(D_METHOD("arcoth_derivv", "z"), &MLPPActivation::arcoth_derivv);
|
|
ClassDB::bind_method(D_METHOD("arcoth_derivm", "z"), &MLPPActivation::arcoth_derivm);
|
|
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_LINEAR);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_SIGMOID);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_SWISH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_MISH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_SIN_C);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_SOFTMAX);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_SOFTPLUS);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_SOFTSIGN);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_ADJ_SOFTMAX);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_C_LOG_LOG);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_LOGIT);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_GAUSSIAN_CDF);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_RELU);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_GELU);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_SIGN);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_UNIT_STEP);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_SINH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_COSH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_TANH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_CSCH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_SECH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_COTH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_ARSINH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_ARCOSH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_ARTANH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_ARCSCH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_ARSECH);
|
|
BIND_ENUM_CONSTANT(ACTIVATION_FUNCTION_ARCOTH);
|
|
}
|