Changelog
0.0.12 (2026-05-02)
Features
- gp: sparse variational markov GP — SparseMarkovGPPrior + SVGP ELBO (#123) (91678ee)
0.0.11 (2026-05-02)
Features
- nn: DenseDVI — Deterministic Variational Inference dense layer (#51) (#134) (7d79679)
- nn: DenseHierarchical — multiplicative shrinkage dense layer (#51) (#132) (265014d)
- nn: LayerNormEnsemble — per-member layer normalisation (#51) (#133) (b880d93)
- nn: MultiHeadAttentionBE — BatchEnsemble multi-head attention (#51) (#135) (8e66cb4)
- nn: output-side noise-contrastive-prior regulariser (#51) (#131) (649ec9b)
0.0.10 (2026-05-02)
Features
0.0.9 (2026-05-02)
Features
- gp: non-Gaussian inference strategies and likelihoods (#47, #48) (#120) (a365252)
- nn: dense variational dropout — sparse variational dropout layer (#51) (#126) (336dbf8)
Bug Fixes
- gp: re-land #114 — SDE composition kernels onto main (#116) (5b36b4d)
- gp: re-land #117 — MarkovGPPrior onto main (orphaned by stacked-PR squash-merge) (#119) (d788dce)
0.0.8 (2026-04-29)
Features
- gp: add SDEKernel protocol and MaternSDE (Markov GP foundation, #37 partial) (#113) (67ce79b)
- nn: canonical cosine RFF layers + spectral-kernel notebook trio (#112) (5840f8c)
0.0.7 (2026-04-28)
Features
- gp: multi-output kernels (LMC/ICM/OILMM) + gaussx-structured inducing operators (#100) (a2451a8)
- gp: pathwise posterior samplers via Matheron's rule (#39) (#101) (714d997)
- nn,gp: spectral layers + inducing features for scalable GPs (#41 + #49) (#84) (c122a87)
- nn: geographic and spherical coordinate encoders (#99) (1709f5f)
- nn: siren — sinusoidal representation network layer + composite (#98) (6f15abd)
- nn: unified conditioning API — FiLM, HyperLinear, HyperSIREN, HyperFourierFeatures (#104) (7232c7c)
0.0.6 (2026-04-21)
Features
- inference: add layered ensemble-of-MAP / ensemble-of-VI runner (#81) (8c44f8f)
- nn: add Bayesian Neural Field layer family, preprocessing, and BNFEstimator (#83) (acbda2c)
0.0.5 (2026-04-16)
Features
- nn: bayesian dense layers, SSGP random features, MC dropout, NCP (#31, #32) (#68) (5e75815)
0.0.4 (2026-04-16)
Features
- gp: natural guide, delta guide, and gaussx-backed guide utilities (#29) (#66) (b709a48)
- gp: svgp inference entry points, CVI, and likelihood families (#30) (#67) (67d80a8)
- gp: wave 3 prior-side whitening + sparse SVGP building blocks (#28) (#64) (4bd48c6)
0.0.3 (2026-04-16)
Features
- gp: wave 2 Epic 2.A — kernel math primitives + Parameterized kernel classes (#60) (f6ee3e1)
- gp: wave 2 Epic 2.B — GPPrior, ConditionedGP, gp_factor, gp_sample (#62) (bd318eb)
0.0.2 (2026-04-16)
Features
- _core: implement PyroxModule, Parameterized, and context lifecycle (#57) (bfdb3d2)
0.0.1 (2026-04-15)
Features
- wave-0: rename scaffold to pyrox and bootstrap package identity (#53) (ea2c95a)
Changelog