2014-08-08 23:04:15 +04:00
|
|
|
#include "dropout_layer.h"
|
2015-03-12 08:20:15 +03:00
|
|
|
#include "params.h"
|
2014-11-19 00:51:04 +03:00
|
|
|
#include "utils.h"
|
2015-01-23 03:38:24 +03:00
|
|
|
#include "cuda.h"
|
2014-11-19 00:51:04 +03:00
|
|
|
#include <stdlib.h>
|
|
|
|
#include <stdio.h>
|
2014-08-08 23:04:15 +04:00
|
|
|
|
2015-05-11 23:46:49 +03:00
|
|
|
dropout_layer make_dropout_layer(int batch, int inputs, float probability)
|
2014-08-08 23:04:15 +04:00
|
|
|
{
|
|
|
|
fprintf(stderr, "Dropout Layer: %d inputs, %f probability\n", inputs, probability);
|
2015-05-11 23:46:49 +03:00
|
|
|
dropout_layer l = {0};
|
|
|
|
l.type = DROPOUT;
|
|
|
|
l.probability = probability;
|
|
|
|
l.inputs = inputs;
|
|
|
|
l.outputs = inputs;
|
|
|
|
l.batch = batch;
|
|
|
|
l.rand = calloc(inputs*batch, sizeof(float));
|
|
|
|
l.scale = 1./(1.-probability);
|
2014-12-13 23:01:21 +03:00
|
|
|
#ifdef GPU
|
2015-05-11 23:46:49 +03:00
|
|
|
l.rand_gpu = cuda_make_array(l.rand, inputs*batch);
|
2014-11-19 00:51:04 +03:00
|
|
|
#endif
|
2015-05-11 23:46:49 +03:00
|
|
|
return l;
|
2014-08-08 23:04:15 +04:00
|
|
|
}
|
|
|
|
|
2015-05-11 23:46:49 +03:00
|
|
|
void resize_dropout_layer(dropout_layer *l, int inputs)
|
2015-02-11 06:41:03 +03:00
|
|
|
{
|
2015-05-11 23:46:49 +03:00
|
|
|
l->rand = realloc(l->rand, l->inputs*l->batch*sizeof(float));
|
2015-02-11 06:41:03 +03:00
|
|
|
#ifdef GPU
|
2015-05-11 23:46:49 +03:00
|
|
|
cuda_free(l->rand_gpu);
|
2015-02-11 06:41:03 +03:00
|
|
|
|
2015-05-11 23:46:49 +03:00
|
|
|
l->rand_gpu = cuda_make_array(l->rand, inputs*l->batch);
|
2015-02-11 06:41:03 +03:00
|
|
|
#endif
|
|
|
|
}
|
|
|
|
|
2015-05-11 23:46:49 +03:00
|
|
|
void forward_dropout_layer(dropout_layer l, network_state state)
|
2014-08-08 23:04:15 +04:00
|
|
|
{
|
|
|
|
int i;
|
2015-03-12 08:20:15 +03:00
|
|
|
if (!state.train) return;
|
2015-05-11 23:46:49 +03:00
|
|
|
for(i = 0; i < l.batch * l.inputs; ++i){
|
2014-12-13 23:01:21 +03:00
|
|
|
float r = rand_uniform();
|
2015-05-11 23:46:49 +03:00
|
|
|
l.rand[i] = r;
|
|
|
|
if(r < l.probability) state.input[i] = 0;
|
|
|
|
else state.input[i] *= l.scale;
|
2014-08-08 23:04:15 +04:00
|
|
|
}
|
|
|
|
}
|
2014-12-13 23:01:21 +03:00
|
|
|
|
2015-05-11 23:46:49 +03:00
|
|
|
void backward_dropout_layer(dropout_layer l, network_state state)
|
2014-08-08 23:04:15 +04:00
|
|
|
{
|
2014-12-13 23:01:21 +03:00
|
|
|
int i;
|
2015-03-12 08:20:15 +03:00
|
|
|
if(!state.delta) return;
|
2015-05-11 23:46:49 +03:00
|
|
|
for(i = 0; i < l.batch * l.inputs; ++i){
|
|
|
|
float r = l.rand[i];
|
|
|
|
if(r < l.probability) state.delta[i] = 0;
|
|
|
|
else state.delta[i] *= l.scale;
|
2014-12-13 23:01:21 +03:00
|
|
|
}
|
2014-08-08 23:04:15 +04:00
|
|
|
}
|
2014-11-19 00:51:04 +03:00
|
|
|
|