# results_graph

**Learned channel-adjacency mixer** on vlaai_ws 10 s (residual A=I+sym(Δ), Δ zero-init). NOT channel-mask+XY FiLM. Swap-test delta=0.0073.

Prep `eaP_n5_t60`, sw2, SPK w10_b3 init (strict=False), max 4 epochs ×250 steps, batch 64. HMM p_stay 0.994 scale 0.8. Also causal hp30.

Code: `graph_adj_clf.py`. Checkpoints: `models/graph_adj/held-*.pt`, `/root/disk_3tb/eeg-thao/models/all5_direct_clf_graphadj_s0.pt`. JSON: `/root/disk_3tb/eeg-thao/results/direct_clf__eaP_n5_t60_w10b3_ptSPK_graphadj.json`.

## LODO (preset)

| Held-out | H | Acc % | hp30 H/Acc | vs stem_ms96 |
|---|---:|---:|---|---|
| AVGC | 0.7073 | 66.35 | 0.6764/63.93 | stem 0.7945/74.06 |
| DTU | 0.8224 | 73.73 | 0.7958/67.25 | stem 0.9061/89.85 |
| KUL | 0.7777 | 71.28 | 0.7771/67.84 | stem 0.8755/82.33 |
| UHD | 0.7516 | 70.17 | 0.7410/66.71 | stem 0.8407/79.47 |
| USTC | 0.7946 | 71.21 | 0.7886/66.75 | stem 0.9219/86.36 |
| **mean** | **0.7707** | **70.55** | **0.7558/66.50** | stem 0.8677/82.42 |

## External (preset)

| Set | H | Acc % | hp30 H/Acc | linear_hp30 | SWA | IRM |
|---|---:|---:|---|---|---|---|
| AASD | 0.6381 | 52.71 | 0.6409/52.81 | 0.6422/53.12 | 0.6202/51.08 | 0.6207/51.01 |
| Carta2026 | 0.6421 | 59.33 | 0.6513/60.46 | 0.6520/60.54 | 0.5507/48.68 | 0.5426/48.04 |

## Targets

| Target | Result |
|---|---|
| AVGC H>=0.85 Acc>=78 | miss (0.7073/66.35) |
| UHD H>=0.85 | miss (0.7516) |
| AASD Acc>60 with H>0.63 | miss (0.6381/52.71; hp30 52.81) |
| LODO mean vs SWA 0.8700/82.61 | 0.7707/70.55 |
| vs IRM AVGC 0.8095/75.12 | AVGC 0.7073/66.35 |
