The detection, modeling and characterization of complex patterns of association in bioinformatics has focused on feature interactions and, more recently, instance-subgroup specific associations (e.g. genetic heterogeneity). Previously, attribute tracking was proposed as an instance-linked memory approach, leveraging the incremental learning of learning classifier systems (LCSs) to track which features were most useful in making class predictions within individual instances. These 'attribute tracking' signatures could later be used to characterize patterns of association in the data. While effective, true underlying patterns remain difficult to characterize in noisy problems, and the original approach places equal weight on tracked feature scores obtained early as well as late in learning. In this work we investigate alternative strategies for attribute tracking scoring, including the adoption of a time recency update scheme taken from reinforcement learning, to gain insight into how to optimize this approach to improve modeling performance and downstream pattern interpretability. We report mixed results over a variety of performance metrics that point to promising future directions for building effective building blocks and improving model interpretability.

Stochastic structured prediction under bandit feedback follows a learning protocol where on each of a sequence of iterations, the learner receives an input, predicts an output structure, and receives partial feedback in form of a task loss evaluation of the predicted structure. We present applications of this learning scenario to convex and non-convex objectives for structured prediction and analyze them as stochastic first-order methods. We present an experimental evaluation on problems of natural language processing over exponential output spaces, and compare convergence speed across different objectives under the practical criterion of optimal task performance on development data and the optimization-theoretic criterion of minimal squared gradient norm. Best results under both criteria are obtained for a non-convex objective for pairwise preference learning under bandit feedback.