diff options
| author | Vasil Zlatanov <v@skozl.com> | 2019-03-07 16:26:28 +0000 |
|---|---|---|
| committer | Vasil Zlatanov <v@skozl.com> | 2019-03-07 16:26:28 +0000 |
| commit | 66ab6413de42d91b8ea5d425636f5721f61b1426 (patch) | |
| tree | 996a824f2215f6f14e4e6cd145aff93c91537bd3 | |
| parent | 23fa20a9a8e8dc34410c400545ef182b0552e72a (diff) | |
| download | e4-gan-66ab6413de42d91b8ea5d425636f5721f61b1426.tar.gz e4-gan-66ab6413de42d91b8ea5d425636f5721f61b1426.tar.bz2 e4-gan-66ab6413de42d91b8ea5d425636f5721f61b1426.zip | |
Add table
| -rw-r--r-- | report/paper.md | 19 |
1 files changed, 19 insertions, 0 deletions
diff --git a/report/paper.md b/report/paper.md index 0227b1e..e1a78c6 100644 --- a/report/paper.md +++ b/report/paper.md @@ -73,6 +73,7 @@ challenge and how they are specifically addressing. Is there the **mode collapse # Inception Score + ## Classifier Architecture Used ## Results @@ -86,6 +87,24 @@ MNIST real testing set (10K), in comparison to the inception scores. **Please measure and discuss the inception scores for the different hyper-parameters/tricks and/or architectures in Q2.** +We measure the performance of the considered GAN's using the Inecption score [@inception], as calculated +with L2-Net logits. + +$$ \textrm{IS}(x) = \exp(\mathcal{E}_x \left( \textrm{KL} ( p(y\|x) \|\| p(y) ) \right) ) $$ + +GAN type Inception Score (L2-Net) +------------ ----------------------------- +MNIST(ref) 9.67 +cGAN 6.01 +cGAN+VB 6.2 +cGAN+LS 6.3 +cGAN+VB+LS 6.4 +cDCGAN+VB 6.5 +cDCGAN+LS 6.8 +cDCGAN+VB+LS 7.3 + + + # Re-training the handwritten digit classifier ## Results |