Outliers with Opposing Signals Have an Outsized Effect on Neural Network Optimization

We identify a new phenomenon in neural network optimization which arises from the interaction of depth and a particular heavy-tailed structure in natural data. Our result offers intuitive explanations for several previously reported observations about network training dynamics, including a conceptually new cause for progressive sharpening and the edge of stability. It also enables new predictions of training behavior which we confirm experimentally, plus a new lens through which to theoretically study and improve modern stochastic optimization on neural nets.Experimentally, we demonstrate the significant influence of paired groups of outliers in the training data with strong \emph{opposing signals}: consistent, large magnitude features which dominate the network output and occur in both groups with similar frequency. Due to these outliers, early optimization enters a narrow valley which carefully balances the opposing groups; subsequent sharpening causes their loss to rise rapidly, oscillating between high on one group and then the other, until the overall loss spikes. We complement these experiments with a theoretical analysis of a two-layer linear network on a simple model of opposing signals.