More Like this

1. A Non-Asymptotic Moreau Envelope Theory for High-Dimensional Generalized Linear Models

   Zhou, L.; Koehler, F.; Sur, P.; Sutherland, D. J.; Srebro, N. (October 2022, A Non-Asymptotic Moreau Envelope Theory for High-Dimensional Generalized Linear Models)

   We prove a new generalization bound that shows for any class of linear predictors in Gaussian space, the Rademacher complexity of the class and the training error under any continuous loss ℓ can control the test error under all Moreau envelopes of the loss ℓ . We use our finite-sample bound to directly recover the “optimistic rate” of Zhou et al. (2021) for linear regression with the square loss, which is known to be tight for minimal ℓ2-norm interpolation, but we also handle more general settings where the label is generated by a potentially misspecified multi-index model. The same argument can analyze noisy interpolation of max-margin classifiers through the squared hinge loss, and establishes consistency results in spiked-covariance settings. More generally, when the loss is only assumed to be Lipschitz, our bound effectively improves Talagrand’s well-known contraction lemma by a factor of two, and we prove uniform convergence of interpolators (Koehler et al. 2021) for all smooth, non-negative losses. Finally, we show that application of our generalization bound using localized Gaussian width will generally be sharp for empirical risk minimizers, establishing a non-asymptotic Moreau envelope theory 

   more » « less
2. Benign Overfitting in Adversarial Training of Neural Networks

   Wang, Yunjuan; Zhang, Kaibo; Arora, Raman (July 2024, Proceedings of the 41st International Conference on Machine Learning, PMLR 235, 2024)

   Benign overfitting is the phenomenon wherein none of the predictors in the hypothesis class can achieve perfect accuracy (i.e., non-realizable or noisy setting), but a model that interpolates the training data still achieves good generalization. A series of recent works aim to understand this phenomenon for regression and classification tasks using linear predictors as well as two-layer neural networks. In this paper, we study such a benign overfitting phenomenon in an adversarial setting. We show that under a distributional assumption, interpolating neural networks found using adversarial training generalize well despite inferencetime attacks. Specifically, we provide convergence and generalization guarantees for adversarial training of two-layer networks (with smooth as well as non-smooth activation functions) showing that under moderate ℓ2 norm perturbation budget, the trained model has near-zero robust training loss and near-optimal robust generalization error. We support our theoretical findings with an empirical study on synthetic and real-world data. 

   more » « less
3. Sublinear Classical and Quantum Algorithms for General Matrix Games

   Li, Tongyang; Wang, Chunhao; Chakrabarti, Shouvanik; Wu, Xiaodi (January 2021, Proceedings of the AAAI Conference on Artificial Intelligence)

   null (Ed.)

   We investigate sublinear classical and quantum algorithms for matrix games, a fundamental problem in optimization and machine learning, with provable guarantees. Given a matrix, sublinear algorithms for the matrix game were previously known only for two special cases: (1) the maximizing vectors live in the L1-norm unit ball, and (2) the minimizing vectors live in either the L1- or the L2-norm unit ball. We give a sublinear classical algorithm that can interpolate smoothly between these two cases: for any fixed q between 1 and 2, we solve, within some additive error, matrix games where the minimizing vectors are in an Lq-norm unit ball. We also provide a corresponding sublinear quantum algorithm that solves the same task with a quadratic improvement in dimensions of the maximizing and minimizing vectors. Both our classical and quantum algorithms are optimal in the dimension parameters up to poly-logarithmic factors. Finally, we propose sublinear classical and quantum algorithms for the approximate Carathéodory problem and the Lq-margin support vector machines as applications. 

   more » « less
4. Heavy Ball Momentum for Conditional Gradient

   Bingcong Li, Alireza Sadeghi (December 2021, Neural Information Processing Systems NeurIPS)

   Conditional gradient, aka Frank Wolfe (FW) algorithms, have well-documented merits in machine learning and signal processing applications. Unlike projection-based methods, momentum cannot improve the convergence rate of FW, in general. This limitation motivates the present work, which deals with heavy ball momentum, and its impact to FW. Specifically, it is established that heavy ball offers a unifying perspective on the primal-dual (PD) convergence, and enjoys a tighter per iteration PD error rate, for multiple choices of step sizes, where PD error can serve as the stopping criterion in practice. In addition, it is asserted that restart, a scheme typically employed jointly with Nesterov’s momentum, can further tighten this PD error bound. Numerical results demonstrate the usefulness of heavy ball momentum in FW iterations. 

   more » « less
5. For interpolating kernel machines, minimizing the norm of the ERM solution maximizes stability

   https://doi.org/10.1142/S0219530522400115  

   Rangamani, Akshay; Rosasco, Lorenzo; Poggio, Tomaso (January 2023, Analysis and Applications)

   In this paper, we study kernel ridge-less regression, including the case of interpolating solutions. We prove that maximizing the leave-one-out ([Formula: see text]) stability minimizes the expected error. Further, we also prove that the minimum norm solution — to which gradient algorithms are known to converge — is the most stable solution. More precisely, we show that the minimum norm interpolating solution minimizes a bound on [Formula: see text] stability, which in turn is controlled by the smallest singular value, hence the condition number, of the empirical kernel matrix. These quantities can be characterized in the asymptotic regime where both the dimension ([Formula: see text]) and cardinality ([Formula: see text]) of the data go to infinity (with [Formula: see text] as [Formula: see text]). Our results suggest that the property of [Formula: see text] stability of the learning algorithm with respect to perturbations of the training set may provide a more general framework than the classical theory of Empirical Risk Minimization (ERM). While ERM was developed to deal with the classical regime in which the architecture of the learning network is fixed and [Formula: see text], the modern regime focuses on interpolating regressors and overparameterized models, when both [Formula: see text] and [Formula: see text] go to infinity. Since the stability framework is known to be equivalent to the classical theory in the classical regime, our results here suggest that it may be interesting to extend it beyond kernel regression to other overparameterized algorithms such as deep networks. 

   more » « less