Research map / Architectures and efficiency / RNNs and normalization
Generalized Shape Metrics on Neural Representations
Alex H Williams, Erin Kunz, Simon Kornblith, Scott Linderman
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 |
|---|---|---|---|
| Replace the mean-canonical-correlation dissimilarity with arccos of the mean correlation | Linear heuristic · Allen Brain Observatory and NAS-Bench-101 Stem layer | triangle inequality violations: 17/1128; 10128/100000 -> 0 by metric construction | Improved |
| Use the shape distance instead of the linear heuristic | Hierarchical clustering of Allen Brain Observatory representations · Allen Brain Observatory | dendrogram similarity: ~85.1% similarity | Mixed |
| Increase the MDS embedding dimension from 2 to a sufficiently large dimension | MDS Euclidean embedding · Allen Brain Observatory and NAS-Bench-101 | pairwise-distance distortion: ~5% (ABO); 10% (NAS-Bench-101) | Improved |
| Use the convolution-aware metric rather than a flattened metric | Convolutional-layer representations · CIFAR-10 test set | convergence of distance estimate with sample size: m=10^4 | Improved |
| Use a looser equivalence relation with more flexible alignments | Generalized shape metrics · Deep convolutional network representations | sample size needed for a stable distance estimate | Hurt |
Closest papers by meaning
- Estimating Shape Distances on Neural Representations with Limited Samples (ICLR 2024 · similarity 0.91)
- Geometry-Aware Similarity Metrics for Neural Representations on Riemannian and Statistical Manifolds (NeurIPS 2026 · similarity 0.89)
- Not all solutions are created equal: An analytical dissociation of functional and representational similarity in deep linear neural networks (ICML 2025 · similarity 0.88)
- Parameter symmetries determine representational geometry in overparameterized nonlinear networks (NeurIPS 2026 · similarity 0.88)
- How Data Augmentation Shapes Neural Representations (NeurIPS 2026 · similarity 0.88)
- Bridging Critical Gaps in Convergent Learning: How Representational Alignment Evolves Across Layers, Training, and Distribution Shifts (NeurIPS 2025 · 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.
