The behavior of the random feature model in the high-dimensional regression framework has become a popular issue of interest in the machine learning literature}. This model is generally considered for feature vectors $x_i = \Sigma^{1/2} x_i'$, where $x_i'$ is a random vector made of independent and identically distributed (iid) entries, and $\Sigma$ is a positive definite matrix representing the covariance of the features. In this paper, we move beyond {\CB this standard assumption by studying the performances of the random features model in the setting of non-iid feature vectors}. Our approach is related to the analysis of the spectrum of large random matrices through random matrix theory (RMT) {\CB and free probability} results. We turn to the analysis of non-iid data by using the notion of variance profile {\CB which} is {\CB well studied in RMT.} Our main contribution is then the study of the limits of the training and {\CB prediction} risks associated to the ridge estimator in the random features model when its dimensions grow. We provide asymptotic equivalents of these risks that capture the behavior of ridge regression with random features in a {\CB high-dimensional} framework. These asymptotic equivalents, {\CB which prove to be sharp in numerical experiments}, are retrieved by adapting, to our setting, established results from operator-valued free probability theory. Moreover, {\CB for various classes of random feature vectors that have not been considered so far in the literature}, our approach allows to show the appearance of the double descent phenomenon when the ridge regularization parameter is small enough.

These lecture notes were developed for a Master's course in advanced machine learning at Erasmus University of Rotterdam. The course is designed for graduate students in mathematics, statistics and econometrics. The content follows a proposition-proof structure, making it suitable for students seeking a formal and rigorous understanding of the statistical theory underlying machine learning methods. At present, the notes focus on linear regression, with an in-depth exploration of the existence, uniqueness, relations, computation, and nonasymptotic properties of the most prominent estimators in this setting: least squares, ridgeless, ridge, and lasso. Background It is assumed that readers have a solid background in calculus, linear algebra, convex analysis, and probability theory.

High-dimensional linear regression has been thoroughly studied in the context of independent and identically distributed data. We propose to investigate high-dimensional regression models for independent but non-identically distributed data. To this end, we suppose that the set of observed predictors (or features) is a random matrix with a variance profile and with dimensions growing at a proportional rate. Assuming a random effect model, we study the predictive risk of the ridge estimator for linear regression with such a variance profile. In this setting, we provide deterministic equivalents of this risk and of the degree of freedom of the ridge estimator. For certain class of variance profile, our work highlights the emergence of the well-known double descent phenomenon in high-dimensional regression for the minimum norm least-squares estimator when the ridge regularization parameter goes to zero. We also exhibit variance profiles for which the shape of this predictive risk differs from double descent. The proofs of our results are based on tools from random matrix theory in the presence of a variance profile that have not been considered so far to study regression models. Numerical experiments are provided to show the accuracy of the aforementioned deterministic equivalents on the computation of the predictive risk of ridge regression. We also investigate the similarities and differences that exist with the standard setting of independent and identically distributed data.

Meta-learning involves training models on a variety of training tasks in a way that enables them to generalize well on new, unseen test tasks. In this work, we consider meta-learning within the framework of high-dimensional multivariate random-effects linear models and study generalized ridge-regression based predictions. The statistical intuition of using generalized ridge regression in this setting is that the covariance structure of the random regression coefficients could be leveraged to make better predictions on new tasks. Accordingly, we first characterize the precise asymptotic behavior of the predictive risk for a new test task when the data dimension grows proportionally to the number of samples per task. We next show that this predictive risk is optimal when the weight matrix in generalized ridge regression is chosen to be the inverse of the covariance matrix of random coefficients. Finally, we propose and analyze an estimator of the inverse covariance matrix of random regression coefficients based on data from the training tasks. As opposed to intractable MLE-type estimators, the proposed estimators could be computed efficiently as they could be obtained by solving (global) geodesically-convex optimization problems. Our analysis and methodology use tools from random matrix theory and Riemannian optimization. Simulation results demonstrate the improved generalization performance of the proposed method on new unseen test tasks within the considered framework.

While the decision-theoretic optimality of the Bayesian formalism under correct model specification is well-known (Berger 2013), the Bayesian case becomes less clear under model misspecification (Grunwald 2017; Ramamoorthi 2015; Fushiki 2005). To formally understand the consequences of Bayesian misspecification, this work examines the relationship between posterior predictive risk and its sensitivity to correct model assumptions, i.e., choice of likelihood and prior. We present the multisample PAC$^m$-Bayes risk. This risk is justified by theoretical analysis based on PAC-Bayes as well as empirical study on a number of toy problems. The PAC$^m$-Bayes risk is appealing in that it entails direct minimization of the Monte-Carlo approximated posterior predictive risk yet recovers both the Bayesian formalism as well as the MLE in its limits. Our work is heavily influenced by Masegosa (2019); our contributions are to align training and generalization risks while offering a tighter bound which empirically performs at least as well and sometimes much better.

Cross-validation is one of the most popular model selection methods in statistics and machine learning. Despite its wide applicability, traditional cross validation methods tend to select overfitting models, due to the ignorance of the uncertainty in the testing sample. We develop a new, statistically principled inference tool based on cross-validation that takes into account the uncertainty in the testing sample. This new method outputs a set of highly competitive candidate models containing the best one with guaranteed probability. As a consequence, our method can achieve consistent variable selection in a classical linear regression setting, for which existing cross-validation methods require unconventional split ratios. When used for regularizing tuning parameter selection, the method can provide a further trade-off between prediction accuracy and model interpretability. We demonstrate the performance of the proposed method in several simulated and real data examples.

To estimate a sparse linear model from data with Gaussian noise, consilience from lasso and compressed sensing literatures is that thresholding estimators like lasso and the Dantzig selector have the ability in some situations to identify with high probability part of the significant covariates asymptotically, and are numerically tractable thanks to convexity. Yet, the selection of a threshold parameter $\lambda$ remains crucial in practice. To that aim we propose Quantile Universal Thresholding, a selection of $\lambda$ at the detection edge. We show with extensive simulations and real data that an excellent compromise between high true positive rate and low false discovery rate is achieved, leading also to good predictive risk.

To what extent are theory-based predictions justified by prior observations? The question is known as the problem of induction and is fundamental to scientific inference. We address the problem of induction from the perspective of learning theory. That is, we consider which theories, and under what assumptions, can be applied to make optimal predictions. Our main result is that the more hypotheses a theory falsifies, suitably quantified, the closer the predictive performance of the best strategy (based on the theory) will be to the theory's post hoc explanatory performance on observed data.