This work concerns learning probabilistic models for ranking data in a heterogeneous population. The specific problem we study is learning the parameters of a Mallows Mixture Model. Despite being widely studied, current heuristics for this problem do not have theoretical guarantees and can get stuck in bad local optima. We present the first polynomial time algorithm which provably learns the pa...

We apply the two-pluyer game assumprio~ls of 1i111ited search horizon and cornn~itnrent to nroves i constant time, to .single-agent heuristic search problems. We present a varicrtion of nrinimcr lookuhead search, and an analog to ulphu-betu pruning rlrot signijicantly improves the efficiency c. the algorithm. Paradoxically. the search horizon reachuble with this algorithm increases wir. increus...

Abstract POPMUSIC is a template for creating heuristics specially designed for dealing with large combinatorial problems that can be partially optimized. The basic idea is to optimize sub-parts of solutions until a local optimum is reached. The useful computational effort should be spent for optimizing these sub-parts. However, it is observed in practice that this computational effort grows alm...

We introduce a flow-dependent version of the quadratic Steiner tree problem in the plane. An instance of the problem on a set of embedded sources and a sink asks for a directed tree T spanning these nodes and a bounded number of Steiner points, such that ∑ e∈E(T ) f(e)|e| is a minimum, where f(e) is the flow on edge e. The edges are uncapacitated and the flows are determined additively, i.e., t...

We investigate optimal designs for discriminating between exponential regression models of different complexity, which are widely used in the biological sciences; see, e.g., Landaw (1995) or Gibaldi and Perrier (1982). We discuss different approaches for the construction of appropriate optimality criteria, and find sharper upper bounds on the number of support points of locally optimal discrimi...

Genetic algorithm (GA) is based on Darwin’s natural selection theory and is used extensively in combinatorial problems as these problems are demanding in terms of computational time. GA shows very good results in terms of both computational time and quality of solution for combinatorial problems as GAs have some traits that make them one of the best evolutionary algorithms (EAs). The use of bot...

We consider the problem of pan-tilt sensor control for active segmentation of incomplete multi-modal data. Since demanding optimal control does not allow for online replanning, we rather employ the optimal planner offline to provide guiding samples for learning a CNN-based control policy in a guided Q-learning framework. The proposed policy initialization and guided Q-learning avoids poor local...