jgrusewski f2938b19e8 fix(ml): improve TPE exploitation with Scott bandwidth, best-trial injection ...

- Replace Silverman's bandwidth (h = 1.06σn^(-1/5)) with Scott's rule
  (h = 0.7σn^(-1/(d+4))) for tighter kernels in high-D parameter spaces
- Add best-trial injection: always evaluate EI at best known point plus
  5 small perturbations (±5%), preventing optimizer from forgetting peaks
- Scale n_candidates dynamically: max(256, 8*n_dims) instead of fixed 100
- Reduce gamma from 0.25 to 0.15 when trials < 50 for tighter exploitation
- Wire model_name through PSO/TPE paths for per-trial Prometheus metrics

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

2026-03-05 23:54:30 +01:00

..

baseline_common

fix(ml): wire spread_cost_bps into RL training — commission + spread = total tx cost

2026-03-01 22:10:43 +01:00

cuda_test.rs

chore: clean up examples, update ML binaries and risk tests

2026-02-27 01:33:18 +01:00

download_baseline.rs

chore: clean up examples, update ML binaries and risk tests

2026-02-27 01:33:18 +01:00

evaluate_baseline.rs

feat(ml): expand DQN hyperopt search space 27D→31D + GPU-batched eval inference

2026-03-05 20:45:13 +01:00

evaluate_supervised.rs

fix(ml): smooth CVaR penalty, fix clip leakage, align noisy sigma, fix eval_supervised

2026-03-05 21:14:31 +01:00

hyperopt_baseline_rl.rs

fix(ml): improve TPE exploitation with Scott bandwidth, best-trial injection

2026-03-05 23:54:30 +01:00

hyperopt_baseline_supervised.rs

fix(ml): improve TPE exploitation with Scott bandwidth, best-trial injection

2026-03-05 23:54:30 +01:00

train_baseline_rl.rs

fix: pass reward-shaping params from hyperopt to walk-forward DQN

2026-03-04 06:24:26 +01:00

train_baseline_supervised.rs

feat(supervised): record learning rate and epoch duration metrics

2026-03-03 00:10:00 +01:00