Hill Space: Neural nets that do perfect arithmetic (to 10⁻¹⁶ precision)

https://news.ycombinator.com/rss Hits: 3
Summary

∞ Unbounded Input Optimizers learn any values they need: -47.2, 156.8, 0.001 → Constraint Function tanh bounds to [-1,1], sigmoid provides gating [-1,1] Bounded Output Maps to [-1,1] range, naturally converging toward discrete selections

First seen: 2025-07-13 08:53

Last seen: 2025-07-13 10:53