1b40fecccd
Implement the Moody & Saffell (2001) DSR for incremental reward shaping that directly optimizes risk-adjusted returns. This replaces the broken double-normalization pipeline (EMA normalizer + risk-adjusted division) that was producing random noise and preventing DQN learning. - DifferentialSharpeRatio struct: step(), reset(), eta clamping, +/-5 bounds - RewardConfig: use_dsr (default false), dsr_eta (default 0.01) - RewardConfigBuilder: use_dsr() and dsr_eta() builder methods - RewardFunction: dsr field, reset_dsr(), reset_epoch_state() - 4 new unit tests (basic, bounded, reset, config roundtrip) - All 40 reward tests pass Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ml
10-model ML ensemble for the Foxhunt HFT system, built on Candle v0.9.1.
Models
- DQN (Rainbow) — deep Q-network with prioritized replay, dueling heads, noisy nets
- PPO — proximal policy optimization with GAE, LSTM policies, clip-higher
- TFT — temporal fusion transformer for multi-horizon forecasting
- Mamba2 — state space model for sequence prediction
- Liquid Networks — biologically inspired networks for non-stationary data
- TLOB — transformer-based limit order book analysis
- KAN — Kolmogorov-Arnold networks
- xLSTM — extended LSTM architecture
- TGGN — temporal graph neural network
- Diffusion — diffusion-based generative model
Key Modules
ensemble— model ensemble coordination and confidence aggregationhyperopt— PSO-based hyperparameter optimization with per-model adapterstrainers— unified training loops (DQN, PPO, supervised)inference—InferenceAdaptertrait for predictioncheckpoint— model checkpointing and restorationevaluation— walk-forward evaluation pipeline
Usage
use ml::dqn::DQN;
use ml::ppo::PpoTrainer;