SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · 2006.05164 · ICML · 2020

AR-DAE: Towards Unbiased Neural Entropy Gradient Estimation

Jae Lim, Aaron Courville, Christopher Pal, Chin-Wei Huang

arXiv · PDF · Open in the Atlas

Code that ran

We lifted 13 functions out of this paper's own repositories and ran 13 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.

RepositoryRoleRan
lim0606/pytorch-ardae-rl — 8 of 8
lim0606/pytorch-ardae-vae — 5 of 5
FunctionStatusWhere it lives
ARDAE Ran lim0606/pytorch-ardae-rl/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("c54ccaa433e557cd")
ARDAE Ran lim0606/pytorch-ardae-vae/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("f2827dbc03163689")
MLP Ran lim0606/pytorch-ardae-rl/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("d795d803d49b3261")
MLP Ran lim0606/pytorch-ardae-vae/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("1bcfbd0bf6acc224")
add_gaussian_noise Ran lim0606/pytorch-ardae-rl/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("227e853b8ec43231")
add_laplace_noise Ran lim0606/pytorch-ardae-vae/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("5750074438e17dce")
add_uniform_noise Ran lim0606/pytorch-ardae-rl/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("647761f0457c5337")
get_nonlinear_func Ran lim0606/pytorch-ardae-rl/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("bba45b256fdad2ad")
grad Ran lim0606/pytorch-ardae-rl/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("6481d3c86e5d4089")
sample_laplace_noise Ran lim0606/pytorch-ardae-vae/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("aa3dcd2b6d66ba3f")
sample_unit_laplace_noise Ran lim0606/pytorch-ardae-vae/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("0acd1ee7ba145eb2")
softplus Ran lim0606/pytorch-ardae-rl/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("813ad67d00d155ae")
swish Ran lim0606/pytorch-ardae-rl/models/graddae/mlp.py
pointer only (licence: NONE) · get_code("0444419acc5bad7c")

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Entropy is ubiquitous in machine learning, but it is in general intractable to compute the entropy of the distribution of an arbitrary continuous random variable. In this paper, we propose the amortized residual denoising autoencoder (AR-DAE) to approximate the gradient of the log density function, which can be used to estimate the gradient of entropy. Amortization allows us to significantly reduce the error of the gradient approximator by approaching asymptotic optimality of a regular DAE, in which case the estimation is in theory unbiased. We conduct theoretical and experimental analyses on the approximation error of the proposed method, as well as extensive studies on heuristics to ensure its robustness. Finally, using the proposed gradient approximator to estimate the gradient of entropy, we demonstrate state-ofthe-art performance on density estimation with variational autoencoders and continuous control with soft actor-critic. {0.001, 0.0001} {0.001, 0.0001} β-annealing no no no {no, 50000} {no, 50000} e-train with train+val no no no no yes Evaluation polyak (decay) -no no no 0.998 polyak (start interation) -no no no {0, 1000, 5000, 10000} neval -40000 40000 20000 20000 Table 8. Hyperparameters for the VAE experiments. toy is the 25 Gaussian dataset. dbmnist and sbmnist are dynamically and statically binarized MNIST, respectively.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("2006.05164")
get_code_for_paper("2006.05164")
have("2006.05164")

Connect an agent — have() is free.