A1615
Title: Recent progress of minimum information dependence modeling
Authors: Keisuke Yano - The Institute of Statistical Mathematics (Japan) [presenting]
Abstract: The Minimum Information Dependency Model is a statistical Model designed to handle higher-order dependencies in mixed-domain data. Applications of the Model are discussed, along with its theoretical properties including its relationship to multimarginal entropic optimal transport. Estimation methods and optimization techniques are presented, including acceleration, stabilization, and uncertainty quantification by combining Adam with the Ruppert-Polyak-Juditsky method.
