SPOCC: Scalable POssibilistic Classifier Combination -- toward robust aggregation of classifiers
This addresses the need for robust classifier combination in machine learning, though it appears incremental as it builds on existing possibility theory frameworks.
The paper tackles the problem of aggregating predictions from multiple classifiers trained on overlapping datasets by proposing a new approach based on possibility theory, which uses adaptive t-norms to handle dependencies and inaccuracies, and proves it has robust aggregation properties.
We investigate a problem in which each member of a group of learners is trained separately to solve the same classification task. Each learner has access to a training dataset (possibly with overlap across learners) but each trained classifier can be evaluated on a validation dataset. We propose a new approach to aggregate the learner predictions in the possibility theory framework. For each classifier prediction, we build a possibility distribution assessing how likely the classifier prediction is correct using frequentist probabilities estimated on the validation set. The possibility distributions are aggregated using an adaptive t-norm that can accommodate dependency and poor accuracy of the classifier predictions. We prove that the proposed approach possesses a number of desirable classifier combination robustness properties.