PV2 ⊧ DaPPer: Data-guided Probabilistic Program synthesizersynthesises probabilistic programs from relational datasets

A tool to synthesize full PPL (probabilistic programming language) models from relational datasets.

Application domain/field

• Probabilistic programming
• Program synthesis

Type of tool

Probabilistic program synthesizer

Expected input

• Dataset
• Causal Order hypothesis (ordering of the dataset column identifiers). This is a hypothesis about the direction of causal links between variables.

Format:

• Dataset: .csv file
• Causal order hypothesis: ?

Expected output

PPL program in a subset of the BLOG language.

Internals

It uses simulated annealing (SA) to synthesize PPL (probabilistic programming language) programs from relational datasets.