Structured Tensor Algebra for Efficient Discrete Probabilistic Inference
summary
An implementation of Buck in Julia called JuliaBux extends the Buck language with external functions and a Julia macro, exposes log‑density and gradients, and makes HMC and NUTS available where they weren’t before. A Gibbs sampler exists but underperforms versus targets, so the roadmap prioritizes high-performance Gibbs, conjugacy analysis, automatic marginalization, and aggressive parallel/GPU compilation (vmap-style) for interoperable, faster discrete integer distribution inference.
AI-written summary · about
related talks
42mBayesDB: Query the Probable Implications of Data
47mApproximate Lifted Inference with Probabilistic Databases
01h 01mBrief Tutorial on Probabilistic Databases
20mGenSQL: A Probabilistic Programming System for Querying Generative Models of Database(…)
41mProbabilistic Data Structures
06mUnlocking graph analytics in DuckDB with SQL/PGQ
17mBring Your Own Data Structures to Datalog
01h 03mQuery Optimization and Evaluation 3: Variable Elimination and Tensor Decomposition