The first framework for bit-identical deep learning training that produces MD5-verified identical weights across independent runs.
March 31, 2026
Original Paper
Bit-Identical Medical Deep Learning via Structured Orthogonal Initialization
arXiv · 2603.28040
The Takeaway
By eliminating randomness in weight initialization, batch ordering, and GPU operations through structured orthogonal basis functions, it solves the 'non-determinism' problem in clinical ML. This is a critical breakthrough for regulatory compliance and auditability in safety-critical AI applications.
From the 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 GP