On the Quantitative Analysis of Decoder-Based Generative Models

November 14, 2016 ยท Entered Twilight ยท ๐Ÿ› International Conference on Learning Representations

๐ŸŒ… TWILIGHT: Old Age
Predates the code-sharing era โ€” a pioneer of its time

"Last commit was 8.0 years ago (โ‰ฅ5 year threshold)"

Evidence collected by the PWNC Scanner

Repo contents: .gitignore, License.txt, README.md, algorithms, experiments, lib, models, sampling

Authors Yuhuai Wu, Yuri Burda, Ruslan Salakhutdinov, Roger Grosse arXiv ID 1611.04273 Category cs.LG: Machine Learning Citations 226 Venue International Conference on Learning Representations Repository https://github.com/tonywu95/eval_gen โญ 130 Last Checked 2 months ago
Abstract
The past several years have seen remarkable progress in generative models which produce convincing samples of images and other modalities. A shared component of many powerful generative models is a decoder network, a parametric deep neural net that defines a generative distribution. Examples include variational autoencoders, generative adversarial networks, and generative moment matching networks. Unfortunately, it can be difficult to quantify the performance of these models because of the intractability of log-likelihood estimation, and inspecting samples can be misleading. We propose to use Annealed Importance Sampling for evaluating log-likelihoods for decoder-based models and validate its accuracy using bidirectional Monte Carlo. The evaluation code is provided at https://github.com/tonywu95/eval_gen. Using this technique, we analyze the performance of decoder-based models, the effectiveness of existing log-likelihood estimators, the degree of overfitting, and the degree to which these models miss important modes of the data distribution.
Community shame:
Not yet rated
Community Contributions

Found the code? Know the venue? Think something is wrong? Let us know!

๐Ÿ“œ Similar Papers

In the same crypt โ€” Machine Learning