NumPyro

E898985

NumPyro is a lightweight probabilistic programming library for Python that leverages JAX to provide high-performance, scalable Bayesian inference with modern MCMC and variational inference algorithms.

Try in SPARQL Jump to: Surface forms Statements Referenced by

All labels observed (1)

Label Occurrences
NumPyro canonical 1

Statements (50)

Predicate Object
instanceOf Python library
probabilistic programming library
basedOn JAX NERFINISHED
compatibleWith JAX NumPy NERFINISHED
JAX random module
designedFor high-performance Bayesian inference
research in probabilistic programming
scalable probabilistic modeling
domain Bayesian statistics
machine learning
probabilistic programming
hasFeature JAX-based random number generation
JIT-compiled log probability evaluation
NumPy-like modeling syntax
automatic differentiation variational inference
diagnostics for MCMC
integration with JAX transformations
model transformations
parallel sampling
plate notation for independence structure
reproducible random seeds
subsampling for large datasets
support for Bayesian neural networks
support for custom distributions
support for custom inference algorithms
support for discrete and continuous distributions
support for hierarchical models
support for probabilistic regression
support for time series models
vectorized MCMC chains
hostedOn GitHub NERFINISHED
inspiredBy Pyro NERFINISHED
license Apache License 2.0
programmingLanguage Python
similarTo PyMC NERFINISHED
Pyro NERFINISHED
TensorFlow Probability NERFINISHED
supports Bayesian inference
GPU acceleration
Hamiltonian Monte Carlo NERFINISHED
Markov chain Monte Carlo NERFINISHED
No-U-Turn Sampler NERFINISHED
TPU acceleration
automatic differentiation
just-in-time compilation
stochastic variational inference
variational inference
vectorized computation
uses XLA compilation via JAX
writtenIn Python NERFINISHED

Referenced by (1)

Full triples — surface form annotated when it differs from this entity's canonical label.