ESM-2
Meta AI 开发的蛋白质语言模型，使用 Transformer 架构在数亿条蛋白质序列上 预训练。该模型学习到的表征包含丰富的进化和结构信息，可用于下游任务 如接触预测、功能预测和结构推断。
| Property | Value |
|---|---|
| Purpose | 基于 Transformer 的蛋白质序列表征学习 |
| Time Complexity | O(n^2 * d) |
| Space Complexity | O(n^2) |
| Year | 2022 |
| Difficulty | Intermediate |
| Languages | Python |
| Category | Protein Language Model |
Complexity Analysis
- Time Complexity:
O(n^2 * d) - Space Complexity:
O(n^2)
Performance Insight: The time complexity of this algorithm is polynomial. High space complexity; consider Hirschberg-style space-optimized variants for very long sequences.
Note: Complexity analysis is based on theoretical models. Actual runtime is affected by input scale, hardware, and implementation optimizations. Benchmark for your specific workload.
Literature & Implementation
Related Tools
ProtTrans · Ankh · ProGen
Tags
language-model transformer representation-learning pretrained