![]() |
Neuron Net
|
Describes the process of gradients calculation for the Neuron Base. More...
Macros | |
#define | def_k_CalcOutputGradient 1 |
Index of Output gradients calculation kernel (CalcOutputGradient) More... | |
#define | def_k_cog_matrix_t 0 |
Target tensor. More... | |
#define | def_k_cog_matrix_o 1 |
Output tensor. More... | |
#define | def_k_cog_matrix_ig 2 |
Tensor of gradients at previous layer. More... | |
#define | def_k_cog_activation 3 |
Activation type (ENUM_ACTIVATION) More... | |
#define | def_k_CalcHiddenGradient 2 |
Index of Hidden gradients calculation kernel (CalcHiddenGradient) More... | |
#define | def_k_chg_matrix_w 0 |
Weights matrix (m+1)*n, where m - number of neurons in previous layer and n - number of neurons in current layer. More... | |
#define | def_k_chg_matrix_g 1 |
Tensor of gradients at current layer. More... | |
#define | def_k_chg_matrix_o 2 |
Output tensor. More... | |
#define | def_k_chg_matrix_ig 3 |
Tensor of gradients at previous layer. More... | |
#define | def_k_chg_outputs 4 |
Number of outputs. More... | |
#define | def_k_chg_activation 5 |
Activation type (ENUM_ACTIVATION) More... | |
Functions | |
__kernel void | CalcOutputGradient (__global double *matrix_t, __global double *matrix_o, __global double *matrix_ig, int activation) |
Describes the process of output gradients calculation for the Neuron Base (CNeuronBaseOCL). More... | |
__kernel void | CalcHiddenGradient (__global double *matrix_w, __global double *matrix_g, __global double *matrix_o, __global double *matrix_ig, int outputs, int activation) |
Describes the process of hidden gradients calculation for the Neuron Base (CNeuronBaseOCL). More... | |
virtual bool | CNeuronBaseOCL::UpdateInputWeights (CObject *SourceObject) |
Dispatch method for defining the subroutine for updating weights. More... | |
virtual bool | CNeuronBaseOCL::calcHiddenGradients (CNeuronBaseOCL *NeuronOCL) |
Method to transfer gradient to previous layer by calling kernel CalcHiddenGradient(). More... | |
virtual bool | CNeuronBaseOCL::calcOutputGradients (CArrayDouble *Target) |
Method of output gradients calculation by calling kernel CalcOutputGradient(). More... | |
Describes the process of gradients calculation for the Neuron Base.
Detailed description on the link.
#define def_k_CalcHiddenGradient 2 |
Index of Hidden gradients calculation kernel (CalcHiddenGradient)
Definition at line 107 of file NeuroNet.mqh.
#define def_k_CalcOutputGradient 1 |
Index of Output gradients calculation kernel (CalcOutputGradient)
Definition at line 101 of file NeuroNet.mqh.
#define def_k_chg_activation 5 |
Activation type (ENUM_ACTIVATION)
Definition at line 113 of file NeuroNet.mqh.
#define def_k_chg_matrix_g 1 |
Tensor of gradients at current layer.
Definition at line 109 of file NeuroNet.mqh.
#define def_k_chg_matrix_ig 3 |
Tensor of gradients at previous layer.
Definition at line 111 of file NeuroNet.mqh.
#define def_k_chg_matrix_o 2 |
Output tensor.
Definition at line 110 of file NeuroNet.mqh.
#define def_k_chg_matrix_w 0 |
Weights matrix (m+1)*n, where m - number of neurons in previous layer and n - number of neurons in current layer.
Definition at line 108 of file NeuroNet.mqh.
#define def_k_chg_outputs 4 |
Number of outputs.
Definition at line 112 of file NeuroNet.mqh.
#define def_k_cog_activation 3 |
Activation type (ENUM_ACTIVATION)
Definition at line 105 of file NeuroNet.mqh.
#define def_k_cog_matrix_ig 2 |
Tensor of gradients at previous layer.
Definition at line 104 of file NeuroNet.mqh.
#define def_k_cog_matrix_o 1 |
Output tensor.
Definition at line 103 of file NeuroNet.mqh.
#define def_k_cog_matrix_t 0 |
Target tensor.
Definition at line 102 of file NeuroNet.mqh.
__kernel void CalcHiddenGradient | ( | __global double * | matrix_w, |
__global double * | matrix_g, | ||
__global double * | matrix_o, | ||
__global double * | matrix_ig, | ||
int | outputs, | ||
int | activation | ||
) |
Describes the process of hidden gradients calculation for the Neuron Base (CNeuronBaseOCL).
Detailed description on the link.
[in] | matrix_w | Weights matrix (m+1)*n, where m - number of neurons in previous layer and n - number of neurons in current layer |
[in] | matrix_g | Tensor of gradients at current layer |
[in] | matrix_o | Previous layer Output tensor |
[out] | matrix_ig | Tensor of gradients at previous layer |
outputs | Number of outputs | |
activation | Activation type (ENUM_ACTIVATION) |
Definition at line 108 of file NeuroNet.cl.
|
virtual |
Method to transfer gradient to previous layer by calling kernel CalcHiddenGradient().
NeuronOCL | Pointer to next layer. |
Definition at line 3204 of file NeuroNet.mqh.
__kernel void CalcOutputGradient | ( | __global double * | matrix_t, |
__global double * | matrix_o, | ||
__global double * | matrix_ig, | ||
int | activation | ||
) |
Describes the process of output gradients calculation for the Neuron Base (CNeuronBaseOCL).
Detailed description on the link.
[in] | matrix_t | Target tensor |
[in] | matrix_o | Output tensor |
[out] | matrix_ig | Tensor of gradients |
activation | Activation type (ENUM_ACTIVATION) |
Definition at line 75 of file NeuroNet.cl.
|
virtual |
Method of output gradients calculation by calling kernel CalcOutputGradient().
Target | Traget value |
Definition at line 3237 of file NeuroNet.mqh.
|
virtual |
Dispatch method for defining the subroutine for updating weights.
SourceObject | Pointer to previos layer. |
Definition at line 3359 of file NeuroNet.mqh.