tfprobability

TensorFlow Probability is an R interface to the TensorFlow Probability Python library, providing tools for statistical computation and probabilistic modeling built on TensorFlow. It enables probabilistic programming and Bayesian inference directly from R.

The package includes distributions for statistical modeling, bijectors (invertible transformations) for normalizing flows, and probabilistic Keras layers for building neural networks with uncertainty quantification. Key capabilities include MCMC and variational inference for Bayesian modeling, hidden Markov models for sequential data, and variational autoencoders with built-in KL divergence handling. It bridges R with TensorFlow’s probabilistic programming ecosystem, making deep probabilistic models accessible to R users.