diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-08-29 18:51:48 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-08-29 18:51:48 -0500 |
| commit | 50fbcfe3644e27eb7352d937633b9de45573062b (patch) | |
| tree | 049ec81e1b41a2f2b2cf87b20c3d7a5c02c6677e | |
| parent | e45bd974e679c9c0636ac4fb0a6c738d582428bc (diff) | |
exp: add clean overclamp to hardware confirmation
| -rw-r--r-- | HARDWARE_CONFIRMATION.md | 4 | ||||
| -rw-r--r-- | experiments/physical_grid_bias_p5.py | 2 |
2 files changed, 4 insertions, 2 deletions
diff --git a/HARDWARE_CONFIRMATION.md b/HARDWARE_CONFIRMATION.md index 1748d85..10d8596 100644 --- a/HARDWARE_CONFIRMATION.md +++ b/HARDWARE_CONFIRMATION.md @@ -18,8 +18,8 @@ experiment. - Published-scale component model: differential gain standard deviation `0.01`, twin input mismatch `0.001 V`, multiplier offset `2.3 V/s`. - Methods: clean, raw imperfection, 16-observation constant calibration, - 16-observation degree-2 SDIL predictor, overclamping, and overclamping plus - the same SDIL predictor. + 16-observation degree-2 SDIL predictor, clean overclamping, imperfect + overclamping, and imperfect overclamping plus the same SDIL predictor. - Standard learning: 600 epochs and `0.001 s` learning exposure per active update. - Overclamping: at most 1,000 epochs, released nudging, `0.0025 s/V`, and the diff --git a/experiments/physical_grid_bias_p5.py b/experiments/physical_grid_bias_p5.py index 4b1e305..7238200 100644 --- a/experiments/physical_grid_bias_p5.py +++ b/experiments/physical_grid_bias_p5.py @@ -38,6 +38,7 @@ METHODS = ( "constant", "sdil", "oracle_neutral", + "overclamp_clean", "overclamp", "overclamp_sdil", ) @@ -275,6 +276,7 @@ def summarize(records: list[dict], methods: tuple[str, ...]) -> dict: ("sdil", "raw"), ("sdil", "constant"), ("sdil", "overclamp"), + ("overclamp_sdil", "overclamp_clean"), ("overclamp_sdil", "overclamp"), ): if first in methods and second in methods: |