Analysis of Generalizability of Deep Neural Networks Based on the Complexity of Decision Boundary
This provides a new quantitative method for assessing generalization in deep learning, addressing a key theoretical gap for researchers and practitioners.
The paper tackles the problem of analyzing the generalizability of deep neural networks by hypothesizing that simpler decision boundaries lead to better performance, and it introduces a Decision Boundary Complexity (DBC) score to measure this, with experiments verifying the hypothesis.
For supervised learning models, the analysis of generalization ability (generalizability) is vital because the generalizability expresses how well a model will perform on unseen data. Traditional generalization methods, such as the VC dimension, do not apply to deep neural network (DNN) models. Thus, new theories to explain the generalizability of DNNs are required. In this study, we hypothesize that the DNN with a simpler decision boundary has better generalizability by the law of parsimony (Occam's Razor). We create the decision boundary complexity (DBC) score to define and measure the complexity of decision boundary of DNNs. The idea of the DBC score is to generate data points (called adversarial examples) on or near the decision boundary. Our new approach then measures the complexity of the boundary using the entropy of eigenvalues of these data. The method works equally well for high-dimensional data. We use training data and the trained model to compute the DBC score. And, the ground truth for model's generalizability is its test accuracy. Experiments based on the DBC score have verified our hypothesis. The DBC is shown to provide an effective method to measure the complexity of a decision boundary and gives a quantitative measure of the generalizability of DNNs.