ruvector/examples/neural-trader/production
Claude b2f55707c7
perf(neural-trader): optimize LSTM, attention, and sentiment
- LSTM: pre-allocate gate vectors, inline sigmoid/tanh (avoid map/reduce)
- MultiHeadAttention: cache-friendly i-k-j matmul, optimized softmax
- FeedForward: pre-allocate hidden layer, manual loops
- LayerNorm: manual mean/variance computation
- Lexicon: char-based word extraction (avoid regex)

Key improvements:
- Buffer push: 1.1M/s (+67%)
- Buffer sample: 319K/s (+22%)
- Lexicon: 346K/s (+16%)
2025-12-31 14:19:27 +00:00
..
drl-portfolio-manager.js feat(neural-trader): add production modules with benchmarks 2025-12-31 14:12:41 +00:00
fractional-kelly.js feat(neural-trader): add production modules with benchmarks 2025-12-31 14:12:41 +00:00
hybrid-lstm-transformer.js perf(neural-trader): optimize LSTM, attention, and sentiment 2025-12-31 14:19:27 +00:00
sentiment-alpha.js perf(neural-trader): optimize LSTM, attention, and sentiment 2025-12-31 14:19:27 +00:00