Figaro is extremely expressive. It can represent a wide variety of models, including:

• Directed and undirected models
• Models in which conditions and constraints are expressed by arbitrary Scala functions
• Models involving inter-related objects
• Open universe models in which the number of objects is unknown
• Models involving discrete and continuous elements
• Models in which the elements are rich data structures, such as trees
• Models with structured decisions
• Models with unknown parameters

Figaro provides a rich library of constructs to build these models, and provides ways to extend this library to create your own model elements.

Figaro’s library of reasoning algorithms is also extensible. Current built-in algorithms include:

• Exact inference using variable elimination
• Belief propagation
• Lazy factored inference for infinite models
• Importance sampling
• Metropolis-Hastings, with an expressive language to define proposal distributions
• Support computation
• Most probable explanation (MPE) using variable elimination or simulated annealing
• Probability of evidence using importance sampling
• Particle Filtering
• Factored frontier
• Parameter learning using expectation maximization
• Gibbs sampling

Figaro provides both regular and anytime reasoning algorithms. In addition to the built-in algorithms, Figaro provides a number of tools to create your own reasoning algorithms.

Figaro can also represent dynamic models and provides a particle filtering algorithm for such models.