tbukuai/patchtst-wavelet-sp500-research
Add TASK_HANDOFF.md for next agent session
Add volatility prediction experiment script
Update AGENT_HANDOFF.md with v6, D1/D2, 12-gap audit, literature crawl, ready-to-run v7 + volatility scripts
Upload v7 training script: log-returns + finance losses
Upload PROJECT_CONCLUSION.md
D2 ckpt: all_done complete
D2 FINAL: returns + walk-forward + baselines
D2 ckpt: baselines_done Baselines computed
D2 ckpt: D2b_done DA=0.4841
D2 ckpt: D2a_done DA=0.5157
Upload PROJECT_CONCLUSION.md
D1a ckpt: all_done D1a-d complete
D1a-d FINAL: close-price experiments complete
D1a ckpt: D1d_done DA=0.5000
D1a ckpt: D1c_done DA=0.4683
D1a ckpt: D1b_done DA=0.4762
D1a ckpt: D1a_done DA=0.5000
Upload notebooks/D2a_returns_baselines.ipynb
Upload notebooks/D1a_paper_exact.ipynb
Delete V7_PLAN.md
Upload GAP_FIX_PLAN.md
D2a: remove v31 from MODEL_REPO and TRACKIO_SPACE, update checkpoint comment to reference D1a
D1a: remove v31 from MODEL_REPO and TRACKIO_SPACE, remove v3/v4 reference in discussion
Upload notebooks/D2a_returns_baselines.ipynb
Delete notebooks/v3.1_xlstm_ts_replication.ipynb
Upload notebooks/v3.1_xlstm_ts_replication.ipynb
Delete train_v7d.py
Delete notebooks/v3.1_xlstm_ts_replication.ipynb, notebooks/v7d_xlstm_ts_replication.ipynb, notebooks/v3.1a_D1a_paper_exact.ipynb, notebooks/v3.1b_D2a_returns_baselines.ipynb
Upload notebooks/D2a_returns_baselines.ipynb
Upload notebooks/D1a_paper_exact.ipynb
Upload notebooks/v3.1b_D2a_returns_baselines.ipynb
Upload notebooks/v3.1a_D1a_paper_exact.ipynb
Upload notebooks/v3.1_xlstm_ts_replication.ipynb
Upload notebooks/v3.1_xlstm_ts_replication.ipynb
Upload train_v7d.py
Upload notebooks/v7d_xlstm_ts_replication.ipynb
Upload V7_PLAN.md
Upload XLSTM_TS_PAPER_BREAKDOWN.md
Upload V7_PLAN.md
Upload GAPS_AND_MISSING_ANALYSIS.md
Update README with v6 results
Update CLAUDE.md with v6 results
Update with v6 results: DA(ctx)=49.6% WF, confirms random
v6 results: DA(ctx)=49.6% WF (20/78 windows), confirms random
v6 ckpt: walkforward WF 25/78 mean=48.9%
v6 ckpt: walkforward WF 20/78 mean=49.6%
v6 ckpt: walkforward WF 15/78 mean=49.2%
v6 ckpt: walkforward WF 9/78 mean=45.5%
v6 ckpt: train_done DA(ctx)=51.9%
v6 ckpt: training ep 56 best=48