Petter Restadh: A Polytopal Method for DAG Structure Learning
Tid: On 2020-12-02 kl 10.15 - 11.15
Abstract: DAG models are used in a variety of areas to model complex systems such those related to genetics and fraud detection. Whilst the models are useful learning a suitable DAG from data is a hard problem. Studený, Hemmecke and Lindner proposed a linear optimization approach to DAG model learning via characteristic imsets and the CIM polytope. We have tried an alternative approach via combinatorially characterizing edges and doing an edge walk. We present the ideas and compare this approach to earlier algorithms.
Zoom meeting ID: 652 1007 4301