Optimization Variance: Exploring Generalization Properties of DNNs

Abstract

Unlike the conventional wisdom in statistical learning theory, the test error of a deep neural network (DNN) often demonstrates double descent: as the model complexity increases, it first follows a classical U-shaped curve and then shows a second descent. Through bias-variance decomposition, recent studies revealed that the bell-shaped variance is the major cause of model-wise double descent (when the DNN is widened gradually). This paper investigates epoch-wise double descent, i.e., the test error of a DNN also shows double descent as the number of training epoches increases. By extending the bias-variance analysis to epoch-wise double descent of the zero-one loss, we surprisingly find that the variance itself, without the bias, varies consistently with the test error. Inspired by this result, we propose a novel metric, optimization variance (OV), to measure the diversity of model updates caused by the stochastic gradients of random training batches drawn in the same iteration. OV can be estimated using samples from the training set only but correlates well with the (unknown) \emph{test} error, and hence early stopping may be achieved without using a validation set.

arXiv commentWork in progress

References in corpus (10)

• Fashion-MNIST: a Novel Image Dataset for Benchmarking Machine Learning Algorithms
• Reconciling modern machine learning practice and the bias-variance trade-off
• Scaling description of generalization with number of parameters in deep learning
• On the Spectral Bias of Neural Networks
• Sharp Minima Can Generalize For Deep Nets
• Rethinking Bias-Variance Trade-off for Generalization of Neural Networks
• Identifying Generalization Properties in Neural Networks
• Rethinking Parameter Counting in Deep Models: Effective Dimensionality Revisited
• Early Stopping in Deep Networks: Double Descent and How to Eliminate it
• Rethink the Connections among Generalization, Memorization and the Spectral Bias of DNNs