Topic: dynamic programming

Three talks examine machine learning within optimization. J. Carlos Castro partitions distributed networks into connected regions sharing an interpretable predictor, contrasting selection of existing models with training on pooled data. A variable-neighbourhood search enforces connectivity; dynamic programming provides exact solutions for particular spatial configurations and a comparison for general networks. Ilaria Ciocci studies nonlinear support-vector machines with hard feature cardinality limits. An exact convex mixed-integer quadratic reformulation uses auxiliary variables for polynomi

Classical models of decision-making, including drift-diffusion and other accumulate-to-bound models, assume that the decision rule for committing to a decision based on accumulated evidence is determined prior to observing the evidence. These predefined rules are typically modeled as thresholds, or bounds, on the accumulated evidence (“decision variable”) that are either fixed [1-2] or collapsing over time [3], in both cases serving to balance decision speed (emphasized when bounds are low and reached quickly) and accuracy (emphasized when bounds are high and allow for more evidence to be accu