Research map / Learning theory and algorithms / Sample complexity
On Error and Compression Rates for Prototype Rules
Omer Kerem, Roi Weiss
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 |
|---|---|---|---|
| use k-NN instead of the prototype methods | notMNIST classification · notMNIST test | error | Improved |
| use 1-NN instead of k-NN | notMNIST classification · notMNIST test | error | Hurt |
| add ProtoComp to 1-NN (ablation) | 1-NN · notMNIST test | compression ratio | Improved |
| add ProtoComp to 1-NN (ablation) | 1-NN · notMNIST test | error | No clear effect |
| use OptiNet or Proto-k-NN rather than 1-NN | notMNIST classification · notMNIST test | error | Improved |
| add ProtoComp to OptiNet (ablation) | OptiNet · notMNIST test | compression ratio | Improved |
| add ProtoComp to OptiNet (ablation) | OptiNet · notMNIST test | error | No clear effect |
| add ProtoComp to Proto-k-NN (ablation) | Proto-k-NN · notMNIST test | compression ratio | Improved |
| add ProtoComp to Proto-k-NN (ablation) | Proto-k-NN · notMNIST test | error | No clear effect |
| use Proto-k-NN + ProtoComp instead of OptiNet + ProtoComp | prototype classification · notMNIST test | compression ratio | Improved |
| use ProtoComp to remove prototypes away from the decision boundary | OptiNet · Euclidean classification under the geometric margin condition | compression rate: n^(-3/(2+d)) log n; n^(-2/(2+d)) | Improved |
| use ProtoComp on an existing prototype classifier (ablation) | prototype rules · Euclidean classification, Lebesgue-almost all inputs | classification error | No clear effect |
Closest papers by meaning
- Universal consistency and rates of convergence of multiclass prototype algorithms in metric spaces (JMLR 2021 · similarity 0.90)
- Supervised learning through the lens of compression (NeurIPS 2016 · similarity 0.87)
- Trading off Consistency and Dimensionality of Convex Surrogates for Multiclass Classification (NeurIPS 2024 · similarity 0.85)
- A Universal Growth Rate for Learning with Smooth Surrogate Losses (NeurIPS 2024 · similarity 0.85)
- Compression of Structured Data with Autoencoders: Provable Benefit of Nonlinearities and Depth (ICML 2024 · similarity 0.85)
- Does Compression Imply Generalization? A Minimum Description Length Perspective (NeurIPS 2026 · 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.
