Files
foxhunt/crates
jgrusewski c2e31e2c40 feat(ml): add multi-timeframe feature fusion (default on, 57→185 dims)
BarResampler aggregates 1m OHLCV to 5m/15m/1h on-the-fly. Four LSTM
encoders (6→64 each) concat to 256, project to 128-dim macro context.
Combined state: 57 position-aware + 128 multi-tf = 185 dims.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
2026-03-03 03:42:50 +01:00
..