Gen.jl

A general-purpose probabilistic programming system with programmable inference, embedded in Julia

• Getting Started
• • Installation
  • Example
• Tutorials and Case Studies

Reference

• Getting Started
• • Installation
  • Example
• Gen.jl
• Choice Maps
• • Mutable Choice Maps
• Generative Function Combinators
• • Map combinator
  • Unfold combinator
  • Recurse combinator
  • Switch combinator
• Probability Distributions
• • Built-In Distributions
  • Defining New Distributions Inline with the @dist DSL
  • Mixture Distribution Constructors
  • Defining New Distributions From Scratch
• Extending Gen
• • Custom gradients
  • Custom incremental computation
  • Custom distributions
  • Custom generative functions
  • Custom modeling languages
• Generative Functions
• • Introduction
  • Mathematical concepts
  • Traces
  • Updating traces
  • Differentiable programming
  • Generative function interface
• Importance Sampling
• Modeling Language Implementation
• • Parsing @gen functions
• Optimizing Trainable Parameters
• Learning Generative Functions
• • Learning from Complete Data
  • Learning from Incomplete Data
  • References
  • API
• MAP Optimization
• Markov chain Monte Carlo (MCMC)
• • MCMC in Gen
  • Built-in Stationary Kernels
  • Enabling Dynamic Checks
  • Composite Kernel DSL
  • Involutive MCMC
  • Reverse Kernels
  • API
• Built-in Modeling Language
• • Annotations
  • Making random choices
  • Calling generative functions
  • Composite addresses
  • Tilde syntax
  • Return value
  • Trainable parameters
  • Differentiable programming
  • Static Modeling Language
• Optimizing Trainable Parameters
• • Parameter update
  • Update configurations
• Particle Filtering
• Selections
• Trace Translators
• • Deterministic Trace Translators
  • General Trace Translators
  • Symmetric Trace Translators
  • Simple Extending Trace Translators
  • Trace Transform DSL
  • API
• Variational Inference
• • Black box variational inference
  • Reparametrization trick
• Tutorials and Case Studies