Projects

Learning Functional Distributions with Private Labels

We study the problem of learning functional distributions in the presence of noise. The functional is a map from features to distributions over a set of labels and is assumed to belong to a known class of hypotheses. Features are generated by a general random process and labels are sampled independently from the feature-dependent distributions and then passed through a noisy kernel. We consider online learning where at each time step a predictor attempts to predict the actual (label) distribution given only the features revealed so far and noisy labels in prior steps.