Schedule

MINDS SEMINAR

MINDS Seminar Series | Miju Ahn (Southern Methodist University) - Tractable continuous approximations in feature and constraint selection problems

MINDS SEMINAR
period : 2022-12-06 ~ 2022-12-06
time : 10:00:00 ~ 11:00:00
개최 장소 : Online streaming (Zoom)
Topic : Tractable continuous approximations in feature and constraint selection problems
개요
Date 2022-12-06 ~ 2022-12-06 Time 10:00:00 ~ 11:00:00
Speaker Miju Ahn Affiliation Southern Methodist University
Place Online streaming (Zoom) Streaming link ID : 688 896 1076 / PW : 54321
Topic Tractable continuous approximations in feature and constraint selection problems
Contents Feature selection is a widely used method in statistical learning to build mathematical models with identified significant features. This presentation introduces two types of problems: a groupwise feature selection and a constraint selection problem via cardinality minimization. In the first problem, we introduce a generalized formulation for groupwise feature selection where subsets of the model variables form non-overlapping groups. We present an alternating direction method of multiplier for the nonconvex formulation and show that the algorithm computes a kind of stationary solution. In the second problem, we study a constraint selection problem where the decision-making involves a set of constraints that is preferred to be satisfied by the model. We formulate the problem as a cardinality minimization problem (CMP) that penalizes the number unsatisfied such soft constraints using an indicator function. Our approach reformulates the discrete CMP as continuous problems. We present an equivalent formulation of a mathematical program with complementarity constraints and an approximation as a difference-of-convex program. The stationary solutions of the alternative formulations are investigated, emphasizing the recovery of the local solutions of the CMP. Our numerical study results demonstrate our method's effectiveness in enforcing desired conditions on several suitable applications.
MinDS MinDS · 2022-10-04 10:26 · Views 545