Variational Lossy Autoencoder
Authors: Xi Chen, Diederik P. Kingma, Tim Salimans, Yan Duan, Prafulla Dhariwal, John Schulman, Ilya Sutskever, Pieter Abbeel
ICLR 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 4 EXPERIMENTS In this paper, we evaluate VLAE on 2D images and leave extensions to other forms of data to future work. ... achieving new state-of-the-art results on MNIST, OMNIGLOT and Caltech-101 Silhouettes density estimation tasks as well as competitive results on CIFAR10. |
| Researcher Affiliation | Collaboration | UC Berkeley, Department of Electrical Engineering and Computer Science Open AI |
| Pseudocode | No | The paper does not contain any section or figure explicitly labeled 'Pseudocode' or 'Algorithm', nor are there any structured, code-like formatted procedures. |
| Open Source Code | No | The paper does not include an explicit statement about releasing the source code for the described methodology, nor does it provide a link to a code repository. |
| Open Datasets | Yes | For evaluation, we use binary image datasets that are commonly used for density estimation tasks: MNIST (Le Cun et al., 1998) (both statically binarized 1 and dynamically binarized version (Burda et al., 2015a)), OMNIGLOT (Lake et al., 2013; Burda et al., 2015a) and Caltech-101 Silhouettes (Marlin et al., 2010). All datasets uniformly consist of 28x28 binary images... In addition to binary image datasets, we have applied VLAE to the CIFAR10 dataset of natural images. |
| Dataset Splits | Yes | Experiments are tuned on the validation set and then final experiment was run with train and validation set, with performance evaluated with test set. |
| Hardware Specification | No | The paper does not provide specific details on the hardware used for running experiments, such as GPU or CPU models, or memory specifications. |
| Software Dependencies | No | The paper mentions software components like TensorFlow, Exponential Linear Units (ELUs), and Adamax, but it does not specify their version numbers, which is required for reproducibility. |
| Experiment Setup | Yes | Appendix A details the experimental setup: 'A latent code of dimension 64 was used. ... We used 4 steps of autoregressive flow and each flow is implemented by a 3-layer MADE that has 640 hidden units and uses Relu... In terms of training, Adamax... was used with a learning rate of 0.002. 0.01 nats/data-dim free bits... Polyak averaging... was used to compute the final parameters, with α = 0.998.' |