As a foundation for inductive inference, Bayesian statistics is attractive because it can provide a complete prescription for learning from data.

As a practical matter many problems that we face today in learning from data, are far more easily solved by Bayesian methods than other approaches. Two phenomena have come together to make this happen:

• The first is the enormous increase in the size of databases, which has brought with it much greater complexity in the data we must analyze; Bayesian statistics, long much more straightforward than other approaches, is able to address, conceptually, the complexity.
  
• The second is tremendous advances in computational methods for Bayesian methods that have made it feasible to compute with large databases.

Our research program in Bayesian methods addresses

• foundations,
• modeling,
• computation, and
• data analysis.