Research map / Probabilistic and scientific modeling / Variational and generative sampling
Infinite Variational Autoencoder for Semi-Supervised Learning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use an infinite mixture of VAEs instead of a single VAE representation (ablation) | SVM classification on Dogs latent representations · Dogs | test accuracy: 49.81\pm1.87 -> 58.1\pm2.63; 63.28\pm0.64 -> 72.28\pm0.2; 74.8\pm0.2 -> 79.8\pm0.18; 79.6\pm0.7 -> 83.9\pm0.24 | Helped |
| Use an infinite mixture of VAEs instead of a single VAE representation (ablation) | SVM classification on ModelNet10 latent representations · ModelNet10 | accuracy: 64.21 -> 74.01; 79.09 -> 83.26; 82.71 -> 85.68 | Helped |
| Increase the number of mixture components from 2 to 10 (ablation) | Infinite Mixture VAE reconstruction · MNIST | reconstruction error: $9.17$ -> $5.12$ | Helped |
| Increase the number of mixture components from 10 to 17 (ablation) | Infinite Mixture VAE reconstruction · MNIST | reconstruction error: $5.12$ -> $4.9$ | Helped |
| Use a mixture of VAEs rather than a single VAE | Semi-supervised image classification · Dogs, 100/1000/4000/all labels | test accuracy: 69.59\pm3.21 -> 75.81\pm1.83; 86.72\pm0.66 -> 89.28\pm0.19; 89.88\pm0.03 -> 90.68\pm0.05; 90.26\pm0.25 -> 91.69\pm0.17 | Helped |
| Use a mixture of VAEs rather than a single ResNet classifier | Semi-supervised image classification · CIFAR-10, 1000/4000 labels | test error: 10.08\pm1.12 -> 8.72\pm0.45; 8.04\pm.21 -> 7.78\pm0.13 | Helped |
| Use a mixture of VAEs rather than a single ResNet classifier | Semi-supervised image classification · CIFAR-10, all labels | test error: $7.5\pm0.01$ -> $7.5\pm0.02$ | No clear effect |
| Increase hidden units in a single VAE from 100 to 1024 (ablation) | VAE reconstruction · MNIST | reconstruction error: $5.92$ -> $5.1$ | Helped |
Closest papers by meaning
- Recursive Inference for Variational Autoencoders (NeurIPS 2020 · similarity 0.90)
- Shape your Space: A Gaussian Mixture Regularization Approach to Deterministic Autoencoders (NeurIPS 2021 · similarity 0.89)
- Hierarchical Decompositional Mixtures of Variational Autoencoders (ICML 2019 · similarity 0.89)
- Consistency Regularization for Variational Auto-Encoders (NeurIPS 2021 · similarity 0.89)
- The Autoencoding Variational Autoencoder (NeurIPS 2020 · similarity 0.88)
- VAE Learning via Stein Variational Gradient Descent (NeurIPS 2017 · similarity 0.88)
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.