augustin64/tipe
1
1
Fork 0
mirror of https://github.com/augustin64/projet-tipe synced 2025-01-23 15:16:26 +01:00
Code Issues Packages Projects Releases Wiki Activity

Fix Adam Optimiser

Browse Source
This commit is contained in:
augustin64 2023-07-12 15:10:29 +02:00
parent b659247283
commit 6f076e67c5
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
src/cnn/update.c
View File

@ -89,7 +89,7 @@ void update_weights(Network* network) {
#ifdef ADAM_DENSE_WEIGHTS
d_nn->v_d_weights[a][b] = BETA_1*d_nn->v_d_weights[a][b] + (1-BETA_1)*d_nn->d_weights[a][b];
d_nn->s_d_weights[a][b] = BETA_2*d_nn->s_d_weights[a][b] + (1-BETA_2)*d_nn->d_weights[a][b]*d_nn->d_weights[a][b];
nn->weights[a][b] -= ALPHA*(d_nn->d_weights[a][b]/sqrt(d_nn->s_d_weights[a][b]+Epsilon));
nn->weights[a][b] -= ALPHA*(d_nn->v_d_weights[a][b]/sqrt(d_nn->s_d_weights[a][b]+Epsilon));
#else
nn->weights[a][b] -= network->learning_rate * d_nn->d_weights[a][b];
#endif
@ -126,8 +126,9 @@ void update_bias(Network* network) {
for (int b=0; b < output_width; b++) {
for (int c=0; c < output_width; c++) {
#ifdef ADAM_CNN_BIAS
d_cnn->v_d_bias[a][b][c] = BETA_1*d_cnn->v_d_bias[a][b][c] + (1-BETA_1)*d_cnn->d_bias[a][b][c];
d_cnn->s_d_bias[a][b][c] = BETA_2*d_cnn->s_d_bias[a][b][c] + (1-BETA_2)*d_cnn->d_bias[a][b][c]*d_cnn->d_bias[a][b][c];
cnn->bias[a][b][c] -= ALPHA*(d_cnn->d_bias[a][b][c]/sqrt(d_cnn->s_d_bias[a][b][c]+Epsilon));
cnn->bias[a][b][c] -= ALPHA*(d_cnn->v_d_bias[a][b][c]/sqrt(d_cnn->s_d_bias[a][b][c]+Epsilon));
#else
cnn->bias[a][b][c] -= network->learning_rate * d_cnn->d_bias[a][b][c];
#endif
@ -148,8 +149,9 @@ void update_bias(Network* network) {
for (int a=0; a < output_width; a++) {
#ifdef ADAM_DENSE_BIAS
d_nn->v_d_bias[a] = BETA_1*d_nn->v_d_bias[a] + (1-BETA_1)*d_nn->d_bias[a];
d_nn->s_d_bias[a] = BETA_2*d_nn->s_d_bias[a] + (1-BETA_2)*d_nn->d_bias[a]*d_nn->d_bias[a];
nn->bias[a] -= ALPHA*(d_nn->d_bias[a]/sqrt(d_nn->s_d_bias[a]+Epsilon));
nn->bias[a] -= ALPHA*(d_nn->v_d_bias[a]/sqrt(d_nn->s_d_bias[a]+Epsilon));
#else
nn->bias[a] -= network->learning_rate * d_nn->d_bias[a];
#endif