Neural Network Subnetworks Are Pure Functions, But Specs Entangle With Data Distribution
A researcher argues that many neural network subnetworks behave like pure functions that could in principle be formally specified, but such specs remain deeply entangled with the training data distribution, making modular verification hard.
2026-11-09 ~ 2026-11-09 · 2 related posts
- Neural net subnetworks are specifiable pure functions — but data entangles the spec — jd_pressman · 2026-11-09
- Neural net subcircuits are pure functions in principle, but their spec is tied to the data distribution — jd_pressman · 2026-11-09