Understanding deep learning (still) requires rethinking generalization

C Zhang, S Bengio, M Hardt, B Recht… - Communications of the …, 2021 - dl.acm.org
Despite their massive size, successful deep artificial neural networks can exhibit a
remarkably small gap between training and test performance. Conventional wisdom …

Fine-grained analysis of optimization and generalization for overparameterized two-layer neural networks

S Arora, S Du, W Hu, Z Li… - … Conference on Machine …, 2019 - proceedings.mlr.press
Recent works have cast some light on the mystery of why deep nets fit any data and
generalize despite being very overparametrized. This paper analyzes training and …

User-friendly introduction to PAC-Bayes bounds

P Alquier - Foundations and Trends® in Machine Learning, 2024 - nowpublishers.com
Aggregated predictors are obtained by making a set of basic predictors vote according to
some weights, that is, to some probability distribution. Randomized predictors are obtained …

Generalization bounds: Perspectives from information theory and PAC-Bayes

F Hellström, G Durisi, B Guedj… - … and Trends® in …, 2025 - nowpublishers.com
A fundamental question in theoretical machine learning is generalization. Over the past
decades, the PAC-Bayesian approach has been established as a flexible framework to …

Leveraging unlabeled data to predict out-of-distribution performance

S Garg, S Balakrishnan, ZC Lipton… - ar** non-vacuous generalization bounds for deep
neural networks, these bounds tend to be uninformative about why deep learning works. In …

Subgroup generalization and fairness of graph neural networks

J Ma, J Deng, Q Mei - Advances in Neural Information …, 2021 - proceedings.neurips.cc
Despite enormous successful applications of graph neural networks (GNNs), theoretical
understanding of their generalization ability, especially for node-level tasks where data are …

Tighter risk certificates for neural networks

M Pérez-Ortiz, O Rivasplata, J Shawe-Taylor… - Journal of Machine …, 2021 - jmlr.org
This paper presents an empirical study regarding training probabilistic neural networks
using training objectives derived from PAC-Bayes bounds. In the context of probabilistic …

What do compressed deep neural networks forget?

S Hooker, A Courville, G Clark, Y Dauphin… - arxiv preprint arxiv …, 2019 - arxiv.org
Deep neural network pruning and quantization techniques have demonstrated it is possible
to achieve high levels of compression with surprisingly little degradation to test set accuracy …