From d9f1b0b16edeb59281355a855e18a8be343fc33c Mon Sep 17 00:00:00 2001
From: Joseph Redmon <pjreddie@gmail.com>
Date: Fri, 08 Aug 2014 19:04:15 +0000
Subject: [PATCH] probably how maxpool layers should be
---
src/connected_layer.c | 111 ++++++++++++++-----------------------------------------
1 files changed, 28 insertions(+), 83 deletions(-)
diff --git a/src/connected_layer.c b/src/connected_layer.c
index 16a39be..368fb63 100644
--- a/src/connected_layer.c
+++ b/src/connected_layer.c
@@ -7,11 +7,16 @@
#include <stdlib.h>
#include <string.h>
-connected_layer *make_connected_layer(int batch, int inputs, int outputs, ACTIVATION activation)
+connected_layer *make_connected_layer(int batch, int inputs, int outputs, ACTIVATION activation, float learning_rate, float momentum, float decay)
{
fprintf(stderr, "Connected Layer: %d inputs, %d outputs\n", inputs, outputs);
int i;
connected_layer *layer = calloc(1, sizeof(connected_layer));
+
+ layer->learning_rate = learning_rate;
+ layer->momentum = momentum;
+ layer->decay = decay;
+
layer->inputs = inputs;
layer->outputs = outputs;
layer->batch=batch;
@@ -24,8 +29,9 @@
layer->weight_momentum = calloc(inputs*outputs, sizeof(float));
layer->weights = calloc(inputs*outputs, sizeof(float));
float scale = 1./inputs;
+ //scale = .01;
for(i = 0; i < inputs*outputs; ++i)
- layer->weights[i] = scale*(rand_uniform());
+ layer->weights[i] = scale*(rand_uniform()-.5);
layer->bias_updates = calloc(outputs, sizeof(float));
layer->bias_adapt = calloc(outputs, sizeof(float));
@@ -39,36 +45,15 @@
return layer;
}
-/*
-void update_connected_layer(connected_layer layer, float step, float momentum, float decay)
+void update_connected_layer(connected_layer layer)
{
int i;
for(i = 0; i < layer.outputs; ++i){
- float delta = layer.bias_updates[i];
- layer.bias_adapt[i] += delta*delta;
- layer.bias_momentum[i] = step/sqrt(layer.bias_adapt[i])*(layer.bias_updates[i]) + momentum*layer.bias_momentum[i];
+ layer.bias_momentum[i] = layer.learning_rate*(layer.bias_updates[i]) + layer.momentum*layer.bias_momentum[i];
layer.biases[i] += layer.bias_momentum[i];
}
for(i = 0; i < layer.outputs*layer.inputs; ++i){
- float delta = layer.weight_updates[i];
- layer.weight_adapt[i] += delta*delta;
- layer.weight_momentum[i] = step/sqrt(layer.weight_adapt[i])*(layer.weight_updates[i] - decay*layer.weights[i]) + momentum*layer.weight_momentum[i];
- layer.weights[i] += layer.weight_momentum[i];
- }
- memset(layer.bias_updates, 0, layer.outputs*sizeof(float));
- memset(layer.weight_updates, 0, layer.outputs*layer.inputs*sizeof(float));
-}
-*/
-
-void update_connected_layer(connected_layer layer, float step, float momentum, float decay)
-{
- int i;
- for(i = 0; i < layer.outputs; ++i){
- layer.bias_momentum[i] = step*(layer.bias_updates[i]) + momentum*layer.bias_momentum[i];
- layer.biases[i] += layer.bias_momentum[i];
- }
- for(i = 0; i < layer.outputs*layer.inputs; ++i){
- layer.weight_momentum[i] = step*(layer.weight_updates[i] - decay*layer.weights[i]) + momentum*layer.weight_momentum[i];
+ layer.weight_momentum[i] = layer.learning_rate*(layer.weight_updates[i] - layer.decay*layer.weights[i]) + layer.momentum*layer.weight_momentum[i];
layer.weights[i] += layer.weight_momentum[i];
}
memset(layer.bias_updates, 0, layer.outputs*sizeof(float));
@@ -78,7 +63,9 @@
void forward_connected_layer(connected_layer layer, float *input)
{
int i;
- memcpy(layer.output, layer.biases, layer.outputs*sizeof(float));
+ for(i = 0; i < layer.batch; ++i){
+ memcpy(layer.output+i*layer.outputs, layer.biases, layer.outputs*sizeof(float));
+ }
int m = layer.batch;
int k = layer.inputs;
int n = layer.outputs;
@@ -86,18 +73,15 @@
float *b = layer.weights;
float *c = layer.output;
gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
- for(i = 0; i < layer.outputs*layer.batch; ++i){
- layer.output[i] = activate(layer.output[i], layer.activation);
- }
- //for(i = 0; i < layer.outputs; ++i) if(i%(layer.outputs/10+1)==0) printf("%f, ", layer.output[i]); printf("\n");
+ activate_array(layer.output, layer.outputs*layer.batch, layer.activation);
}
-void learn_connected_layer(connected_layer layer, float *input)
+void backward_connected_layer(connected_layer layer, float *input, float *delta)
{
int i;
for(i = 0; i < layer.outputs*layer.batch; ++i){
layer.delta[i] *= gradient(layer.output[i], layer.activation);
- layer.bias_updates[i%layer.batch] += layer.delta[i]/layer.batch;
+ layer.bias_updates[i%layer.outputs] += layer.delta[i];
}
int m = layer.inputs;
int k = layer.batch;
@@ -105,55 +89,16 @@
float *a = input;
float *b = layer.delta;
float *c = layer.weight_updates;
- gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
+ gemm(1,0,m,n,k,1,a,m,b,n,1,c,n);
+
+ m = layer.batch;
+ k = layer.outputs;
+ n = layer.inputs;
+
+ a = layer.delta;
+ b = layer.weights;
+ c = delta;
+
+ if(c) gemm(0,1,m,n,k,1,a,k,b,k,0,c,n);
}
-void backward_connected_layer(connected_layer layer, float *input, float *delta)
-{
- memset(delta, 0, layer.inputs*sizeof(float));
-
- int m = layer.inputs;
- int k = layer.outputs;
- int n = layer.batch;
-
- float *a = layer.weights;
- float *b = layer.delta;
- float *c = delta;
-
- gemm(0,0,m,n,k,1,a,k,b,n,1,c,n);
-}
-/*
- void forward_connected_layer(connected_layer layer, float *input)
- {
- int i, j;
- for(i = 0; i < layer.outputs; ++i){
- layer.output[i] = layer.biases[i];
- for(j = 0; j < layer.inputs; ++j){
- layer.output[i] += input[j]*layer.weights[i*layer.inputs + j];
- }
- layer.output[i] = activate(layer.output[i], layer.activation);
- }
- }
- void learn_connected_layer(connected_layer layer, float *input)
- {
- int i, j;
- for(i = 0; i < layer.outputs; ++i){
- layer.delta[i] *= gradient(layer.output[i], layer.activation);
- layer.bias_updates[i] += layer.delta[i];
- for(j = 0; j < layer.inputs; ++j){
- layer.weight_updates[i*layer.inputs + j] += layer.delta[i]*input[j];
- }
- }
- }
- void backward_connected_layer(connected_layer layer, float *input, float *delta)
- {
- int i, j;
-
- for(j = 0; j < layer.inputs; ++j){
- delta[j] = 0;
- for(i = 0; i < layer.outputs; ++i){
- delta[j] += layer.delta[i]*layer.weights[i*layer.inputs + j];
- }
- }
- }
- */
--
Gitblit v1.10.0