Research map / Architectures and efficiency / RNNs and normalization
Diet Networks: Thin Parameters for Fat Genomics
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 reconstruction loss to Diet Networks with per-class histogram embeddings (ablation) | Diet Networks (per-class histograms) · 1000 Genomes ancestry prediction | misclassification error (%): 7.88 ± 1.40 -> 7.44 ± 0.45 | Helped |
| Add reconstruction loss to Diet Networks with random-projection embeddings (ablation) | Diet Networks (random projection) · 1000 Genomes ancestry prediction | misclassification error (%): 9.03 ± 1.20 -> 8.03 ± 1.03 | Helped |
| Add reconstruction loss to Diet Networks with raw end-to-end embeddings (ablation) | Diet Networks (raw end-to-end) · 1000 Genomes ancestry prediction | misclassification error (%): 8.88 ± 1.42 -> 8.28 ± 1.92 | Helped |
| Add reconstruction loss to Diet Networks with SNP2Vec embeddings (ablation) | Diet Networks (SNP2Vec) · 1000 Genomes ancestry prediction | misclassification error (%): 7.60 ± 1.28 -> 7.88 ± 0.72 | Hurt |
| Use raw end-to-end feature embeddings instead of the basic network | Diet Networks · 1000 Genomes ancestry prediction | misclassification error (%): 8.31 ± 1.83 -> 8.88 ± 1.42 | Hurt |
| Use precomputed SNP2Vec embeddings instead of random-projection embeddings | Diet Networks · 1000 Genomes ancestry prediction | misclassification error (%): 9.03 ± 1.20 -> 7.60 ± 1.28 | Helped |
| Use per-class histogram embeddings instead of random-projection embeddings, with reconstruction | Diet Networks with reconstruction · 1000 Genomes ancestry prediction | misclassification error (%): 8.03 ± 1.03 -> 7.44 ± 0.45 | Helped |
| Increase the PCA feature count from 10 to 200 | PCA classifier · 1000 Genomes ancestry prediction | misclassification error (%): 20.56 ± 3.20 -> 9.33 ± 1.24 | Helped |
| Add hidden layers to a classifier using PCA features (ablation) | PCA (100 PCs) classifier · 1000 Genomes ancestry prediction | misclassification error (%): 10.52 ± 0.25 -> 12.67 ± 0.67; 12.18 ± 1.75; 11.95 ± 2.29 | Hurt |
| Increase the PCA feature count beyond 200 (ablation) | PCA classifier · 1000 Genomes ancestry prediction | misclassification error (%) | Hurt |
Closest papers by meaning
- Training Flexible Models of Genetic Variant Effects from Functional Annotations using Accelerated Linear Algebra (ICML 2025 · similarity 0.85)
- Weight Predictor Network with Feature Selection for Small Sample Tabular Biomedical Data (AAAI 2023 · similarity 0.84)
- GV-Rep: A Large-Scale Dataset for Genetic Variant Representation Learning (NeurIPS 2024 · similarity 0.84)
- AffinityNet: Semi-Supervised Few-Shot Learning for Disease Type Prediction (AAAI 2019 · similarity 0.84)
- GeSubNet: Gene Interaction Inference for Disease Subtype Network Generation (ICLR 2025 · similarity 0.83)
- The Effect of Intrinsic Dataset Properties on Generalization: Unraveling Learning Differences Between Natural and Medical Images (ICLR 2024 · similarity 0.82)
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.