Aki Vehtari

Overview

Aki Vehtari is a professor of computational probabilistic modeling at Aalto University (Finland), a core contributor to Bayesian model assessment and to the Stan ecosystem. He is known for practical methods for predictive model evaluation — cross-validation, WAIC, and Pareto-smoothed importance sampling.

Key Contributions / Features

• PSIS-LOO — Vehtari–Gelman–Gabry (2017): With Andrew Gelman and Jonah Gabry, fast and stable leave-one-out cross-validation via Pareto-smoothed importance sampling (PSIS), with the $\hat k$ reliability diagnostic and the R loo package. See WAIC, Vehtari-Gelman-Gabry (2017), and Importance Sampling.
• Survey of predictive methods (Vehtari–Ojanen 2012): A comprehensive review of Bayesian predictive model assessment and selection.
• Contributor to the loo, brms, and projpred R packages and to Bayesian workflow practice.

Related

• Widely Applicable Information Criterion
• Importance Sampling
• Model Comparison
• Andrew Gelman
• Jonah Gabry
• Vehtari-Gelman-Gabry (2017)