Research map / Learning theory and algorithms / Sample complexity
Conservativeness of Untied Auto-Encoders
Read the paperCheck 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 |
|---|---|---|---|
| Add weight length constraints (ablation) | MNIST auto-encoders · MNIST | symmetricity of ADW: 95.9% -> 98.7%; 95.1% -> 99.1% | Helped |
| Add weight length constraints (ablation) | MNIST contractive auto-encoders · MNIST | symmetricity of ADW: 95.2% -> 98.6%; 97.4% -> 99.1% | Helped |
| Use contractive training instead of unregularized AE training (ablation) | MNIST untied auto-encoder · MNIST | symmetricity of ADW: 95.9% -> 95.2%; 95.1% -> 97.4% | Mixed |
| Increase beta in the mixture toward the conservative field | Tied AE conservative-component extraction · MNIST / corrupted MNIST | fraction with E(x) > E(x_rand): 0.5036 -> 0.9968 | Helped |
| Increase beta in the mixture toward the conservative field | Untied AE conservative-component extraction · MNIST / corrupted MNIST | fraction with E(x) > E(x_rand): 0.5072 -> 0.9968 | Helped |
Closest papers by meaning
- Compression of Structured Data with Autoencoders: Provable Benefit of Nonlinearities and Depth (ICML 2024 · similarity 0.85)
- Laplacian Autoencoders for Learning Stochastic Representations (NeurIPS 2022 · similarity 0.85)
- High-dimensional Asymptotics of Denoising Autoencoders (NeurIPS 2023 · similarity 0.85)
- Exact Rate-Distortion in Autoencoders via Echo Noise (NeurIPS 2019 · similarity 0.84)
- Fundamental Limits of Two-layer Autoencoders, and Achieving Them with Gradient Methods (ICML 2023 · similarity 0.84)
- On Random Deep Weight-Tied Autoencoders: Exact Asymptotic Analysis, Phase Transitions, and Implications to Training (ICLR 2019 · similarity 0.84)
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.