Fluctuations, bias, variance and ensemble of learners: exact asymptotics for convex losses in high-dimension

From the sampling of data to the initialisation of parameters, randomness is ubiquitous in modern Machine Learning practice. Understanding the statistical fluctuations engendered by the different sources of randomness in prediction is therefore key to understanding robust generalisation. In this manuscript we develop a quantitative and rigorous theory for the study of fluctuations in an ensemble of generalised linear models trained on different, but correlated, features in high-dimensions. In particular, we provide a complete description of the asymptotic joint distribution of the empirical risk minimiser for generic convex loss and regularisation in the high-dimensional limit. Our result encompasses a rich set of classification and regression tasks, such as the lazy regime of overparametrised neural networks, or equivalently the random features approximation of kernels. While allowing to study directly the mitigating effect of ensembling (or bagging) on the bias-variance decomposition of the test error, our analysis also helps disentangle the contribution of statistical fluctuations, and the singular role played by the interpolation threshold that are at the roots of the 'double-descent' phenomenon.

Automated summary

This manuscript develops a quantitative and rigorous theory for studying the fluctuations in an ensemble of generalized linear models trained on high-dimensional correlated features. The results can be applied to various classification and regression tasks and help understand the impact of ensembling on test error as well as the roots of the "double-descent" phenomenon.