Release notes
This is the 0.20 release of TensorFlow Probability. It is
tested and stable against TensorFlow version 2.12 and JAX 0.4.8 .
Change notes
- Add `LinearOperatorBasis` and `LinearOperatorRowBlock`.
- Ensure `Dirichlet` and `RelaxedOneHotCategorical` transform correctly under bijectors.
- Add `SphericalSpace` and use in all Spherical Distributions
- Add `GeneralSpace.transform_general`
- Fix guitar numpy rewrite_equivalence_test.
- BREAKING CHANGE: Ignore deprecated `always_yield_multivariante_normal` arg to `tfd.GaussianProcess` and `tfd.GaussianProcessRegressionModel` so that event shape is always [1] for a single index point.
- Create a `bayesopt` submodule of TFP experimental and add acquisition functions.
- Add the `FeatureScaledWithCategorical` kernel, a PSD kernel over structures of continuous and categorical data, to TFP experimental.
- [BREAKING] Remove deprecated arg BDF.use_pfor_to_compute_jacobian.
Huge thanks to all the contributors to this release!
- ashishenoy
- atondwal
- bjp
- Christopher Suter
- colcarroll
- Colin Carroll
- emilyaf
- fdtomasi
- flang
- Jacob Burnim
- jburnim
- jcater
- juanantoniomc
- langmore
- Leandro Campos
- leben
- Matthew Feickert
- mmladenov
- nkovela
- Pavel Sountsov
- phandu
- phawkins
- power
- S. Amin
- siege
- Srinivas Vasudevan
- synandi
- thomaswc
- Tirumalesh
- ujaved
- ursk