nx_ng_hashenc.nx

nx_ng_hashenc.nx -- CAP-HASH-ENCODING, generation 9: the core contribution of Instant-NGP (Mueller et al. 2022, arxiv 2201.05989, the repo the operator linked). The measured CLAIM that justifies the method: a LEARNABLE interpolated FEATURE-GRID encoding represents high-frequency detail that a model of the RAW coordinate cannot -- cheap, fast, "seconds not hours". Honest measured-exceed, isolated on the ENCODING (operator: capability is the exceed, measured, no wave). Two models trained on the SAME maximally-high-frequency target, differing ONLY in how the input coordinate is represented: PLAIN out_i = a*x_i + b -- the raw coordinate (affine). Provably can only draw a LINE, so it cannot fit an oscillating signal. (convex -> trains robustly to its best-line plateau = an honest, non-fragile failure.) HASH out_i = interp(feat[i],feat[i+1]) -- a learnable interpolated feature grid (the encoding core). Linear in the features -> convex -> fits the high-freq field. Gate: T1 HASH FITS (SSE collapses >=95%), T2 PLAIN FAILS (its SSE stays high), T3 MEASURED EXCEED (loss_plain >= 4x loss_hash), T4 BIT-EXACT (train twice -> identical integer features). HONEST: this isolates the encoding; composing it with a trainable MLP readout (proven in nx_ng_radiance) + the MULTI-resolution levels + spatial hashing (3D/high-dim scaling) are the follow-on rungs. Sovereign: nx_nofloat_autograd + syscalls, no float. license_tier: ORIGINAL expect_exit: 0