Event

We usually design materials to target desired behaviors that are defined in a top-down manner. Learning theory offers an alternative when desired behaviors are hard to define but easy to give examples of. We consider materials that change as they physically experiences training examples and then test the material on novel inputs never seen before (generalization).  We study the physical requirements for such information processing using theory and experiments in two systems: thin sheets that can be trained in a supervised manner to classify high dimensional force patterns and self-assembling DNA strands that can classify high dimensional patterns in chemical concentrations.