Slowing Learning by Erasing Simple Features
This work addresses the fundamental learning dynamics in neural networks, but it is incremental as it builds on prior erasure methods to explore specific effects on learning speed.
The authors tackled the problem of how neural networks learn by removing low-order statistical information from data, finding that quadratic erasure (QLEACE) consistently slows learning in feedforward architectures, while approximate variants can enhance learning speed on some datasets.
Prior work suggests that neural networks tend to learn low-order moments of the data distribution first, before moving on to higher-order correlations. In this work, we derive a novel closed-form concept erasure method, QLEACE, which surgically removes all quadratically available information about a concept from a representation. Through comparisons with linear erasure (LEACE) and two approximate forms of quadratic erasure, we explore whether networks can still learn when low-order statistics are removed from image classification datasets. We find that while LEACE consistently slows learning, quadratic erasure can exhibit both positive and negative effects on learning speed depending on the choice of dataset, model architecture, and erasure method. Use of QLEACE consistently slows learning in feedforward architectures, but more sophisticated architectures learn to use injected higher order Shannon information about class labels. Its approximate variants avoid injecting information, but surprisingly act as data augmentation techniques on some datasets, enhancing learning speed compared to LEACE.