Files

jgrusewski 2df1ea92e1 feat(ml): WAVE 29 DQN Codebase Cleanup & Refactoring Campaign

BREAKING CHANGES:
- Removed orphaned dqn.rs monolithic trainer (4,975 lines)
- Removed orphaned dqn_ensemble.rs module (816 lines)
- Removed orphaned tft.rs and tft_complete_int8_integration_test.rs
- TFT trainer split into modular directory structure

DQN Module Refactoring:
- Split trainers/dqn.rs into modular structure (config.rs, statistics.rs, trainer.rs)
- Fixed hyperopt 39D search space (continuous params only)
- Boolean flags (use_dueling, use_double_dqn, use_per, use_noisy_nets) are now FIXED architectural decisions
- use_distributional defaults to false (Candle BUG #36 - scatter_add gradient issues)

Clean Module Structure:
- ml/src/trainers/dqn/ directory with proper mod.rs exports
- ml/src/trainers/tft/ directory with config.rs, types.rs, model.rs, trainer.rs, tests.rs
- All P0 features validated: TD-error clamping, batch diversity, LR scheduler, priority staleness

Documentation:
- Added comprehensive docs in docs/codebase-cleanup/
- ADR-001 for DQN refactoring decisions
- Rainbow DQN component matrix and quick reference guides

Build Status: Compiles with zero errors

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <noreply@anthropic.com>

2025-11-27 23:46:13 +01:00

1.6 KiB

Raw Blame History

DQN Backtesting Report

Model: DQN-New-Model Baseline: DQN-Trial35-Baseline Generated: 2025-11-04 07:56:33 UTC

Performance Summary

Metric	Value	Target	Status
Total Return	18.50%	>0%	✅
Sharpe Ratio	2.50	>1.5	✅
Max Drawdown	10.20%	<20%	✅
Win Rate	62.0%	>50%	✅
Alpha vs B&H	4.50%	>0%	✅

Comparison to Baseline

Metric	Baseline	New Model	Change	Direction
Returns	12.10%	18.50%	+6.40%	↗️
Sharpe	1.80	2.50	+0.70	↗️
Drawdown	18.30%	10.20%	-8.10%	↗️
Win Rate	52.0%	62.0%	+10.0%	↗️
Alpha	1.50%	4.50%	+3.00%	↗️

Trade Statistics

Metric	Value
Total Trades	150
Avg Trade Return	0.123%
Win Rate	62.00%
Trades vs Baseline	+12

Deployment Recommendation

Status: ✅ APPROVE - Ready for Production

Model passes 5/5 production criteria. Strong performance with 18.50% return, 2.50 Sharpe ratio, and 62.0% win rate. Risk is acceptable with 10.20% max drawdown. Model demonstrates profitability with 4.50% alpha vs buy-and-hold.

Action: Proceed with production deployment after final validation.

Production Criteria Checklist

Criteria Passed: 5/5
Total Return: ✅ PASS (18.50% > 0%)
Sharpe Ratio: ✅ PASS (2.50 > 1.5)
Max Drawdown: ✅ PASS (10.20% < 20%)
Win Rate: ✅ PASS (62.0% > 50%)
Alpha vs B&H: ✅ PASS (4.50% > 0%)

Report generated automatically by Foxhunt ML Evaluation Framework

1.6 KiB Raw Blame History