AI Paper Atlas

Research map / Probabilistic and scientific modeling / Variational and generative sampling

Pythae: Unifying Generative Autoencoders in Python - A Benchmarking Use Case

Clément Chadebec, Louis J. Vincent, Stephanie Allassonniere

NeurIPS 2022 · Probabilistic and scientific modeling › Variational and generative sampling · approach: Hierarchical VAEs and posterior collapse · 6 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Fit a 10-component GMM on latent codes instead of sampling the N(0,I) prior (ablation)AE · MNIST ConvNet generationFID: 26.7 -> 9.3Improved
Fit a GMM instead of N(0,I) prior sampling (ablation)beta-VAE · MNIST ConvNet generationFID: 21.4 -> 9.2Improved
Fit a GMM instead of N(0,I) prior sampling (ablation)VAE · MNIST ConvNet generationFID: 28.5 -> 26.9Improved
Replace GMM with MAF normalising flow sampler (ablation)VAE · MNIST ConvNet generationFID: 26.9 -> 26.8No clear effect
Use 2-stage VAE sampler instead of N(0,I) prior (ablation)VAE · MNIST ConvNet generationFID: 28.5 -> 40.3Hurt
Use VAMP prior instead of standard Gaussian priorVAE · MNIST ConvNet generationFID: 28.5 -> 64.2Hurt
Use small-beta beta-VAE instead of standard VAEVAE · MNIST ConvNet reconstructionMSE (x1e-3): 16.85 -> 5.61Improved
Add L2 penalty on decoder weights (RAE-L2) to AEAE · CIFAR10 ConvNet reconstructionMSE (x1e-3): 2.82 -> 2.25Improved
Use beta-TC VAE instead of standard VAEVAE · CIFAR10 ConvNet classificationaccuracy: 32.61 -> 45.40Improved
Use autoencoder instead of VAEVAE · MNIST ConvNet classificationaccuracy: 86.75 -> 93.81Improved
Use more flexible VAMP prior instead of standard priorVAE · MNIST ConvNet classificationaccuracy: 86.75 -> 92.17Improved
Use VAEGAN adversarial reconstruction loss instead of MSE lossVAE · MNIST ConvNet reconstructionFID: 30.71 -> 5.54Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy