Contact weight ridge preserved beta SEEG fits across three equivalent difference bases in 19 selected records

Complete adjacent and anchor differences contain the same within-block signal information. We tested whether a ridge penalty preserves SEEG coupling after this change of basis. We used 19 selected interictal records from OpenNeuro ds003876 version 1.0.2. The beta band was 13 to 29.5 Hz. Each fit used 30 two-second training segments and 30 next-minute evaluation segments. The primary ridge strength was alpha 0.1. Channel ridge changed ordinary squared coherence by a mean absolute value of 0.034136. Its fitted leading block pair changed in 9 records. A penalty on contact weights preserved fitted and evaluation scores across three bases. Its largest primary evaluation difference was 2.302e-14. A proof requires complete equal row spaces. Single-mode predictions also require an isolated leading value. All tested leading gaps were positive. The package contains inputs, code, weights, source receipts, and reproduction instructions. Fresh source downloads rebuilt all inputs exactly at complex64 precision. A full offline repeat matched 61 output files byte for byte. This method test does not establish neural connections or clinical accuracy.

stereoelectroencephalography · SEEG · regularized canonical correlation · ridge regression · difference basis · contact weights · representation sensitivity · beta band