A learned state reduced mean 7 Hz filter phase error after 100 ms simulated gaps in nine SEEG rest records with known initial states

1. usesA shared reader had 58.38 degrees of mean absolute phase error in two new SEEG rest records at 50 ms with a centered theta reference sci:sha256:6bd2f301da5cc69ca85251358ec5d79565bf3c48c78038666bfdd36352965b56

   This publication supplied all nine selected voltage extracts, corrected contact mapping, and inherited source provenance. New public prefix downloads reproduced every selected voltage value.

2. usesA 7 Hz causal filter reader reduced unseen input squared error by a mean 50.4 percent at 50 ms in nine SEEG rest records sci:sha256:f851bd1de5c8367288ad9745a4316baf6f4e7fde8dfec5bb56d0adb133df2ac4

   This publication supplied the causal 7 Hz filter design, known versus missing input split, 27 past input features, and chronological ridge fitting design. The new study tested recovery after input returned and preserved an added last sample test.

3. usesTwo exact bounds certified a 7 Hz SEEG filter phase after simulated gaps in nine rest records under voltage caps sci:sha256:1e7436d8e30edf812519ca308906d17bd0b1a757df88c2ccc6c9d70a987f5cb5

   The public source reconstruction recipe was adapted from this publication. Its gap phase bounds also informed the distinction between mean recovery error and an individual phase guarantee. The new learned reader has no error certificate.

4. usesdoi:10.18112/openneuro.ds006065.v1.0.0

   The study used selected rest recordings from ds006065 version 1.0.0, as stated in the source table; the versioned metadata records CC0.

5. usesdoi:10.18112/openneuro.ds005670.v1.0.0

   The study used selected rest recordings from ds005670 version 1.0.0, as stated in the source table; the versioned metadata records CC0.

Public catalog entries on Sciscape.

• usesA reader of past voltage reduced 7 Hz filter phase error in nine SEEG rest records with known initial state and 100 ms simulated gaps every 200 ms

  This study uses the missing filter input target, early calibration splits, ridge penalty grid, and fixed 7 Hz causal filter. It tests a new map with voltage history features and no state features.

• usesA constrained fit reduced mean 7 Hz filter phase error against a guard applied after fitting in nine SEEG records with simulated mixed gaps and known initial states

  This test uses the original past feature and ridge fitting code. It retains the fitting cutoffs and selected penalties.

• usesTwo of 36 learned SEEG state maps were unstable during 100 ms simulated gaps every 120 ms

  We reused all nine selected native voltage extracts, fixed 100 ms gap models, causal filter code, public-source reconstruction code, and source provenance. We replaced isolated exact-state replays with continuous repeated-gap replays and tested state feedback stability.

• usesA 0.95 gap map bound limited state sensitivity but increased phase error under dense irregular loss in nine SEEG rest records

  Supplied all 144 fixed models and their earlier fitting recipe. This replay uses the same weights and scales. A separate check rebuilt the weights.

• supportsA learned state with the last voltage sample reduced mean 7 Hz filter phase error against held voltage after 20 ms simulated gaps in 36 SEEG patients with exact pre-gap states

  The exploratory last-voltage feature had lower mean 20 ms gap recovery error than held voltage and the 27-feature model in 36 selected ds004100 patients. The held-voltage reduction was 0.54 degrees, and 31 of 36 patients improved. This supports the feature in this new dataset under exact pre-gap states. It does not reproduce the same cohort or establish live stimulation performance.

• usesA learned state with the last voltage sample reduced mean 7 Hz filter phase error against held voltage after 20 ms simulated gaps in 36 SEEG patients with exact pre-gap states

  We adapted the causal gap reader, missing-input regression target, 27-feature model, exploratory last-voltage feature, and recovery identity. We tested both feature sets in a different public SEEG dataset with file A fitting and file B tests.

• usesAlternating stable schedules made two of 36 fixed SEEG readers unstable during simulated 100 ms gaps

  This study supplied the fixed 27-feature learned readers and their early fitting design. We use its weights as fixed artifacts and do not fit new models.

• usesWith shared input history, singular value clipping bounded state sensitivity in 36 fixed 100 ms SEEG readers

  The fixed 27-feature model weights, calibration, and fitting times came from this isolated-gap study. We did not fit the models again.

• usesThe phase range midpoint reduced the first certified wait for a 7 Hz filter after simulated gaps in nine SEEG rest records with known states and voltage limits

  Supplied all nine exact public rest voltage extracts, fixed pair selection, corrected EDF mapping, source metadata, licenses, and the source recovery recipe. No learned model or model score from that work entered this analysis.

• usesIn nine SEEG rest records, a 0.95 gap-map bound limited state growth but raised phase error under dense irregular loss

  Supplied all 144 fixed models and their earlier fitting recipe. This replay uses the same weights and scales. A separate check rebuilt the weights.

• usesA new fit of voltage history weights reduced state error after map clipping in nine SEEG rest records with sparse simulated loss

  We used its 144 original models, selected penalties, feature means, scales, and target means. We rebuilt its unconstrained weights from the early voltage data. New fits kept its model coordinates and penalties.

• usesAlternating two stable timing maps made one fixed SEEG state reader unstable with 100 ms simulated gaps

  This earlier work supplied the 27-feature learned gap models and causal filter utilities inherited through the repeated-gap parent. We preserved the fixed weights and early calibration without fitting new models.

• usesA learned state reduced mean 7 Hz filter phase error after repeated 100 ms gaps in nine SEEG records with unchanged returned voltage

  This archive supplied the nine selected voltage files, both fixed model files, filter code, fitting history, and source reconstruction code. The new test retains earlier state and buffer errors across repeated gaps.

• usesProducts of two stable cycle maps were unstable in 11 of 36 fitted SEEG readers on a fixed timing grid

  This study used the fixed planned 27-feature gap models and their early voltage calibration. No weights were refit and no contacts were selected again.

• usesA reader of past voltage reduced mean 7 Hz filter phase error in nine SEEG rest records with mixed 20 to 200 ms simulated gaps and known initial states

  This original nine-record study supplied the selected pairs and early voltage calibration in the inherited research line. We retained those choices to compare readers on the same voltage and filter target.

• usesFitting under a 0.95 state norm limit reduced mean 7 Hz filter phase error by 14.26 degrees against a scaled guard in nine SEEG records with changing simulated gaps of 20 to 200 ms

  Its original affine regression code and early fitting plan define the target, features, and inherited validation-selected penalties. We reconstructed all 144 original fits exactly and fit new weights under a complete gap map norm limit.

• usesA fixed learned update reduced 7 Hz filter phase error in nine SEEG rest records with simulated 100 ms gaps every 250 ms

  This experiment uses its fixed 100 ms gap models, filter code, voltage normalization values, nine input arrays, and selected pairs. It replaces isolated gaps with continuous repeated-gap replay and retained state and history errors.