Research map / Label-efficient and robust learning / Recognition and long-tail
Estimating Information Flow in Deep Neural Networks
Ziv Goldfeld, Ewout Van Den Berg, Kristjan Greenewald, Igor Melnyk, Nam Nguyen, Brian Kingsbury, Yury Polyanskiy
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Increase additive-noise variance from β=0.005 to a higher-variance setting | SZT model · SZT | mutual information compression | Hurt |
| Use ReLU instead of tanh | SZT model · SZT | mutual information compression | Hurt |
| Add orthonormal weight regularization to the ReLU network (ablation) | SZT model · SZT | mutual information compression | Hurt |
| Use a deterministic network and vary the bin size used for MI estimation | SZT model · SZT | estimated mutual information | Mixed |
| Cluster inputs using a non-saturating Leaky-ReLU network | Two-neuron Leaky-ReLU network · Synthetic two-class example | mutual information | Mixed |
| Use ReLU without weight normalization (ablation) | SZT model · SZT | mutual information compression | Hurt |
| Use ReLU with orthonormal weight regularization (ablation) | SZT model · SZT | mutual information compression | Hurt |
Closest papers by meaning
- Adaptive Estimators Show Information Compression in Deep Neural Networks (ICLR 2019 · similarity 0.90)
- Information Bottleneck Analysis of Deep Neural Networks via Lossy Compression (ICLR 2024 · similarity 0.89)
- Information Bottleneck: Exact Analysis of (Quantized) Neural Networks (ICLR 2022 · similarity 0.88)
- InfoNet: Neural Estimation of Mutual Information without Test-Time Optimization (ICML 2024 · similarity 0.86)
- Does Compression Imply Generalization? A Minimum Description Length Perspective (NeurIPS 2026 · similarity 0.85)
- Understanding and Leveraging the Learning Phases of Neural Networks (AAAI 2024 · similarity 0.85)
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.
