Research map / Architectures and efficiency / RNNs and normalization
Recurrent Kernel Networks
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 |
|---|---|---|---|
| Use RKN instead of CKN with BLOSUM62 encoding and max pooling | Protein fold recognition · SCOP 1.67 | auROC: 0.866 -> 0.871 | Helped |
| Use RKN instead of CKN with BLOSUM62 encoding and max pooling | Protein fold recognition · SCOP 1.67 | auROC50: 0.621 -> 0.629 | Helped |
| Use RKN instead of CKN with 512 filters | Protein fold classification · SCOP 2.06 | fold-level top1 accuracy: 45.41 -> 71.35 | Helped |
| Use RKN instead of CKN with 512 filters | Protein fold classification · SCOP 2.06 | top 1 accuracy: 84.11 -> 85.29 | Helped |
| Add gap-aware matching (RKN rather than λ=0 RKN) (ablation) | RKN, BLOSUM62 max pooling · SCOP 1.67 | auROC: 0.862 -> 0.871 | Helped |
| Add gap-aware matching (RKN rather than λ=0 RKN) (ablation) | RKN, BLOSUM62 max pooling · SCOP 1.67 | auROC50: 0.618 -> 0.629 | Helped |
| Use max pooling instead of mean pooling (ablation) | RKN, BLOSUM62 · SCOP 1.67 | auROC: 0.840 -> 0.871 | Helped |
| Use max pooling instead of mean pooling (ablation) | RKN, BLOSUM62 · SCOP 1.67 | auROC50: 0.571 -> 0.629 | Helped |
| Use GMP pooling instead of max pooling (ablation) | RKN, one-hot encoding · SCOP 1.67 | auROC: 0.844 -> 0.848 | Helped |
| Use GMP pooling instead of max pooling (ablation) | RKN, BLOSUM62 · SCOP 1.67 | auROC: 0.871 -> 0.852 | Hurt |
| Use max pooling instead of mean pooling (ablation) | Unsupervised RKN, one-hot encoding · SCOP 1.67 | auROC: 0.805 -> 0.801 | Hurt |
| Use the summed-kernel RKN variant instead of single-k RKN (ablation) | Unsupervised RKN, one-hot encoding, mean pooling · SCOP 1.67 | auROC: 0.805 -> 0.821 | Helped |
Closest papers by meaning
- Kernel-Based Approaches for Sequence Modeling: Connections to Neural Methods (NeurIPS 2019 · similarity 0.89)
- Deriving Neural Architectures from Sequence and Graph Kernels (ICML 2017 · similarity 0.87)
- Parallelizing Linear Recurrent Neural Nets Over Sequence Length (ICLR 2018 · similarity 0.87)
- Learning and Generalization in RNNs (NeurIPS 2021 · similarity 0.86)
- Universality of Linear Recurrences Followed by Non-linear Projections: Finite-Width Guarantees and Benefits of Complex Eigenvalues (ICML 2024 · similarity 0.85)
- A Tensor Decomposition Perspective on Second-order RNNs (ICML 2024 · 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.