Work completed at the University of Washington. Work completed at Microsoft Research.

We study the Kronecker product regression problem, in which the design matrix is a Kronecker product of two or more matrices.

We develop a general framework for finding approximately-optimal preconditioners for solving linear systems.

Leslie Valiant's 1984 paper "A Theory of the Learnable" [Val84], reproduced in this volume, has the unusual distinction of having changed the course of several scientific disciplines. Within theoretical computer science it was one of the key works giving rise to the field now known as computational learning theory, which may loosely be defined as the rigorous study of learning processes and phenomena from the computer science perspective of efficient algorithms and computational complexity. In the decades since the publication of [Val84], computational learning theory has grown into a rich field with strong connections to many other theoretical disciplines such as mathematical probability and statistics, information theory, decision theory and more. Beyond the realm of theory, Valiant's paper and the Probably Approximately Correct (PAC) model which he introduced in it have also had a great impact on the subsequent development of machine learning, a field which has already transformed many aspects of science and human society and seems certain to have an even greater influence in the future. This chapter gives an overview of the Probably Approximately Correct learning model that Valiant introduced in [Val84], explaining some of the major results and directions that the field has taken in the years since that work.

Recent studies suggest that asymmetric binary perceptron (ABP) likely exhibits the so-called statistical-computational gap characterized with the appearance of two phase transitioning constraint density thresholds: \textbf{\emph{(i)}} the \emph{satisfiability threshold} $α_c$, below/above which ABP succeeds/fails to operate as a storage memory; and \textbf{\emph{(ii)}} \emph{algorithmic threshold} $α_a$, below/above which one can/cannot efficiently determine ABP's weight so that it operates as a storage memory. We consider a particular parametric utilization of \emph{fully lifted random duality theory} (fl RDT) [85] and study its potential ABP's algorithmic implications. A remarkable structural parametric change is uncovered as one progresses through fl RDT lifting levels. On the first two levels, the so-called $\c$ sequence -- a key parametric fl RDT component -- is of the (natural) decreasing type. A change of such phenomenology on higher levels is then connected to the $α_c$ -- $α_a$ threshold change. Namely, on the second level concrete numerical values give for the critical constraint density $α=α_c\approx 0.8331$. While progressing through higher levels decreases this estimate, already on the fifth level we observe a satisfactory level of convergence and obtain $α\approx 0.7764$. This allows to draw two striking parallels: \textbf{\emph{(i)}} the obtained constraint density estimate is in a remarkable agrement with range $α\in (0.77,0.78)$ of clustering defragmentation (believed to be responsible for failure of locally improving algorithms) [17,88]; and \textbf{\emph{(ii)}} the observed change of $\c$ sequence phenomenology closely matches the one of the negative Hopfield model for which the existence of efficient algorithms that closely approach similar type of threshold has been demonstrated recently [87].

We develop a general framework for finding approximately-optimal preconditioners for solving linear systems.

Leverage score sampling is a powerful technique that originates from theoretical computer science, which can be used to speed up a large number of fundamental questions, e.g.