a24fd7c9bd
- iql_value_kernel.cu: remove old per-sample kernels (iql_forward_loss_kernel, iql_backward_per_sample, iql_weight_grad_reduce) replaced by cuBLAS path - gpu_experience_collector.rs: delete _portfolio_dim dead variable - metrics.rs: replace hardcoded feature_dim=62 with market_dim+OFI_DIM - config.rs: bars_per_day default 390.0→0.0, add validation at training start - training_loop.rs: fail fast if bars_per_day==0 (uninitialized) - common_device_functions.cuh: remove unused PORTFOLIO_DIM define, update comments to "42 market + 14 portfolio", keep STATE_DIM (used in TILE_LAYER_WARP_CLEAN) - experience_kernels.cu: remove unused PORTFOLIO_DIM define - dqn.rs, config.rs: update stale "42 market + 8 portfolio" comments to 14 Co-Authored-By: Claude Sonnet 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;