[2603.28040] Bit-Identical Medical Deep Learning via Structured Orthogonal Initialization

Computer Science > Machine Learning arXiv:2603.28040 (cs) [Submitted on 30 Mar 2026] Title:Bit-Identical Medical Deep Learning via Structured Orthogonal Initialization Authors:Yakov Pyotr Shkolnikov View a PDF of the paper titled Bit-Identical Medical Deep Learning via Structured Orthogonal Initialization, by Yakov Pyotr Shkolnikov View PDF HTML (experimental) Abstract:Deep learning training is non-deterministic: identical code with different random seeds produces models that agree on aggregate metrics but disagree on individual predictions, with per-class AUC swings exceeding 20 percentage points on rare clinical classes. We present a framework for verified bit-identical training that eliminates three sources of randomness: weight initialization (via structured orthogonal basis functions), batch ordering (via golden ratio scheduling), and non-deterministic GPU operations (via architecture selection and custom autograd). The pipeline produces MD5-verified identical trained weights across independent runs. On PTB-XL ECG rhythm classification, structured initialization significantly exceeds Kaiming across two architectures (n=20; Conformer p = 0.016, Baseline p < 0.001), reducing aggregate variance by 2-3x and reducing per-class variability on rare rhythms by up to 7.5x (TRIGU range: 4.1pp vs 30.9pp under Kaiming, independently confirmed by 3-fold CV). A four-basis comparison at n=20 shows all structured orthogonal bases produce equivalent performance (Friedman p=0.48), esta...