Active Learning under distributional shift

 Personnel: Mohit Prabhushankar

Goal: To construct Active Learning algorithms that are generalizable to noise and distributional shifts


Challenges: Active Learning query strategies are designed to estimate the best subset of data for labeling, that allows a neural network to learn efficiently. However, the central premise of such strategies is that the deployed test set is distributionally similar to the training set. If there is a distributional difference, then the networks fail. While this is true for all machine learning algorithms, Active Learning strategies are especially fallible to this distributional shift.

Our Work: All active learning query strategies operate by estimating the uncertainty with respect to a given sample. For robust uncertainty estimation, we advocate for two stages in a neural network's decision making process. The first is the existing feed-forward inference framework where patterns in given data are sensed and associated with previously learned patterns. The second stage is a slower reflection stage where we ask the network to reflect on its feed-forward decision by considering and evaluating all available choices. Together, we term the two stages as introspective learning. We use gradients of trained neural networks as a measurement of this reflection. A simple three-layered Multi Layer Perceptron is used as the second stage that predicts based on all extracted gradient features. We show that when train and test distributions are dissimilar including during the presence of noise, introspective networks provide accuracy gains in an active learning setting on 5 query strategies. 

Extensions: We perceptually visualize the post-hoc explanations from both stages to provide a visual grounding to introspection. For the application of recognition, we show that an introspective network is 4% more robust and 42% less prone to calibration errors when generalizing to noisy data. We also illustrate the value of introspective networks in downstream tasks that require generalizability and calibration including active learning, out-of-distribution detection, and uncertainty estimation. We ground the proposed machine introspection to human introspection for the application of image quality assessment.

References

  1. M. Prabhushankar and G. AlRegib, "Introspective Learning : A Two-Stage Approach for Inference in Neural Networks" Advances in Neural Information Processing Systems (2022), Nov 29 - Dec 1, 2022.