Generalization Comparison of Deep Neural Networks via Output Sensitivity

Abstract

Although recent works have brought some insights into the performance improvement of techniques used in state-of-the-art deep-learning models, more work is needed to understand their generalization properties. We shed light on this matter by linking the loss function to the output's sensitivity to its input. We find a rather strong empirical relation between the output sensitivity and the variance in the bias-variance decomposition of the loss function, which hints on using sensitivity as a metric for comparing the generalization performance of networks, without requiring labeled data. We find that sensitivity is decreased by applying popular methods which improve the generalization performance of the model, such as (1) using a deep network rather than a wide one, (2) adding convolutional layers to baseline classifiers instead of adding fully-connected layers, (3) using batch normalization, dropout and max-pooling, and (4) applying parameter initialization techniques.

An Adaptive Intrusion Detection System in the Internet of Medical Things Using Fuzzy-Based Learning

“…Moreover, the importance of early stopping in addressing overfitting is emphasized in the literature. Forouzesh and Salehi [40] mention early stopping as one of the regularization techniques applied to avoid overfitting in deep learning architectures. Additionally, Choi and Lee [41] propose a learning strategy that involves training all samples with good initialization parameters and stopping the model using early stopping techniques to prevent overfitting.…”