-
Your `Progressive Differentiable Architecture Search: Bridging the Depth Gap between Search and Evaluation` is interesting! May I suggest one more reference "Searching for A Robust Neural Architecture…
D-X-Y updated
5 years ago
-
Hi! Thanks for your good job!
I have some questions about the Fast Acceleration version. In the main paper, it is said to adopt Gumbel Softmax with temperature to relax argmax function during the s…
-
In the `gumbel_softmax_vae_v2` notebook, for calculating the KL in case of Relaxed Prior, shouldn't the KL be `y * (log q_y - log p_y)`. So the product with `y` is missing?
-
Many people are actually based on the code written in this paper(https://arxiv.org/pdf/1611.01144.pdf), but none of them implement the implementation part of Bernoulli variables. According to https://…
-
Hello and congratulations for your great work.
I tried to implement the L2X explanations with hierarchical LSTM on a fake news dataset, as shown in the imdb_sent example. However, the explainer alway…
-
## Description
Implement biasd-path reparametrization instead of learning-signal based optimization for learning the controller.
Differentiable relaxation of the sampling provide a more controllab…
-
`torch.multinomial` throws a `RuntimeError` if the probability distribution includes `float('inf')`. On CPU there is no similar problem. This can happen if you take a (what is probably naive) approach…
-
### 指摘事項
SampleLayerのノイズの計算で下記の式となっていますが、
これだとノイズが大きいと思ったのですが、どのような意図でこの式となっているのでしょうか?
`return tf.argmax(x - tf.log(-tf.log(noise)), axis=1)`
### 指摘箇所
* [ ] Day4: 強化学習に対するニューラルネットワークの適用
ページ番…
-
Is it ok to directly use argmax rather than gumbel softmax?
-
- [ ] Don't do the ADVI approach. Using priors and Q in same family, will yield closed form KL(q(z)||p(z)). Avoid the jacobians.
- [ ] [Gumbel softmax](https://github.com/dev4488/VAE_gumble_softmax/b…