Adversarial Variational Bayes: Unifying Variational Autoencoders and Generative Adversarial Networks

Authors: Lars Mescheder, Sebastian Nowozin, Andreas Geiger

ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We empirically demonstrate that our model is able to learn rich posterior distributions and show that the model is able to generate compelling samples for complex data sets.
Researcher Affiliation Collaboration 1Autonomous Vision Group, MPI T ubingen 2Microsoft Research Cambridge 3Computer Vision and Geometry Group, ETH Z urich. Correspondence to: Lars Mescheder <lars.mescheder@tuebingen.mpg.de>.
Pseudocode Yes Algorithm 1 Adversarial Variational Bayes (AVB)
Open Source Code No The paper does not provide any links to source code or state that code is made available.
Open Datasets Yes We applied this to the Eight School example from Gelman et al. (2014). [...] In addition, we trained deep convolutional networks based on the DC-GAN-architecture (Radford et al., 2015) on the binarized MNIST-dataset (Le Cun et al., 1998). An additional experiment on the celeb A dataset (Liu et al., 2015) can be found in the Supplementary Material.
Dataset Splits No The paper mentions a test set size for MNIST but does not provide complete training/validation/test splits, percentages, or refer to a standard split that includes all three for any dataset.
Hardware Specification No The paper does not explicitly describe the hardware used for running its experiments.
Software Dependencies No The paper mentions tools like STAN and ITE-package, but does not provide specific version numbers for these or other general software dependencies (e.g., Python, deep learning frameworks) used in their experimental setup.
Experiment Setup Yes Both the encoder and decoder are parameterized by 2-layer fully connected neural networks with 512 hidden units each. [...] The adversary is parameterized by two neural networks with two 512-dimensional hidden layers each [...]. [...] For the decoder network, we use a 5-layer deep convolutional neural network. [...] For the adversary, we replace the fully connected neural network acting on z and x with a fully connected 4-layer neural networks with 1024 units in each hidden layer. [...] For every posterior update step we performed two steps for the adversary.