Probabilistic Decoupling of Labels in Classification
This provides a general solution for classification problems like semi-supervised and noisy-label learning, but it is incremental as it builds on existing probabilistic frameworks.
The paper tackles the problem of decoupling training labels from underlying classes to enable a general inference scheme for various classification tasks, achieving competitive performance on Fashion MNIST and 20 News Groups datasets with simulated noise and partial labeling.
We investigate probabilistic decoupling of labels supplied for training, from the underlying classes for prediction. Decoupling enables an inference scheme general enough to implement many classification problems, including supervised, semi-supervised, positive-unlabelled, noisy-label and suggests a general solution to the multi-positive-unlabelled learning problem. We test the method on the Fashion MNIST and 20 News Groups datasets for performance benchmarks, where we simulate noise, partial labelling etc.