Research map / Label-efficient and robust learning / Recognition and long-tail
Neural Distance Embeddings for Biological Sequences
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 hyperbolic distance instead of best competing geometry for the embedding space (ablation) | NeuroSEED convolutional encoders · edit distance approximation (Qiita, RT988, Greengenes) | % RMSE: 22% average RMSE reduction | Helped |
| use data-dependent neural encoders instead of data-independent k-mer/FFP baselines (ablation) | edit distance approximation · Qiita, RT988, Greengenes | % RMSE | Helped |
| use data-dependent neural encoders on synthetic random sequences (ablation) | edit distance approximation · synthetic random sequences | % RMSE | No clear effect |
| use hyperbolic space for lower embedding dimension (ablation) | NeuroSEED · Qiita edit distance approximation | % RMSE at fixed dimension: matches other spaces at 128 with only 4 to 16 dimensions | Helped |
| run agglomerative clustering on pretrained hyperbolic NeuroSEED embeddings instead of ground-truth distances | hierarchical clustering · Qiita 10k | Dasgupta's cost | No clear effect |
| use more complex encoders (MLP, CNN) instead of a linear layer in relaxed Dasgupta approach (ablation) | NeuroSEED hierarchical clustering · RT988, Qiita | Dasgupta's cost | No clear effect |
| build Clustal guide tree from attention-model NeuroSEED embeddings instead of real distances | multiple sequence alignment · RT988 1200 sequences | alignment cost: 35 minutes tree construction replaced | No clear effect |
| decode Steiner string from geometric median of embeddings versus centre string | NeuroSEED autoencoder · Steiner string consensus error | average consensus error | Helped |
Closest papers by meaning
- Neural Embeddings for kNN Search in Biological Sequence (AAAI 2024 · similarity 0.90)
- Hyperbolic Genome Embeddings (ICLR 2025 · similarity 0.87)
- Levenshtein Distance Embedding with Poisson Regression for DNA Storage (AAAI 2024 · similarity 0.86)
- When is an Embedding Model More Promising than Another? (NeurIPS 2024 · similarity 0.86)
- What DNA Foundation Models Learn Beyond Sequence Composition (NeurIPS 2026 · similarity 0.84)
- Training Compute-Optimal Protein Language Models (NeurIPS 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.