Generalization Bound and Learning Methods for Data-Driven Projections in Linear Programming
This work addresses the fundamental challenge of scaling linear programming for practitioners in optimization and data science, offering a novel approach with theoretical guarantees and practical improvements.
The paper tackles the problem of solving high-dimensional linear programs efficiently by proposing data-driven projections to reduce dimensionality, achieving significantly higher solution quality and faster solving times compared to random projections.
How to solve high-dimensional linear programs (LPs) efficiently is a fundamental question. Recently, there has been a surge of interest in reducing LP sizes using random projections, which can accelerate solving LPs independently of improving LP solvers. This paper explores a new direction of data-driven projections, which use projection matrices learned from data instead of random projection matrices. Given training data of $n$-dimensional LPs, we learn an $n\times k$ projection matrix with $n > k$. When addressing a future LP instance, we reduce its dimensionality from $n$ to $k$ via the learned projection matrix, solve the resulting LP to obtain a $k$-dimensional solution, and apply the learned matrix to it to recover an $n$-dimensional solution. On the theoretical side, a natural question is: how much data is sufficient to ensure the quality of recovered solutions? We address this question based on the framework of data-driven algorithm design, which connects the amount of data sufficient for establishing generalization bounds to the pseudo-dimension of performance metrics. We obtain an $\tilde{\mathrm{O}}(nk^2)$ upper bound on the pseudo-dimension, where $\tilde{\mathrm{O}}$ compresses logarithmic factors. We also provide an $Ω(nk)$ lower bound, implying our result is tight up to an $\tilde{\mathrm{O}}(k)$ factor. On the practical side, we explore two simple methods for learning projection matrices: PCA- and gradient-based methods. While the former is relatively efficient, the latter can sometimes achieve better solution quality. Experiments demonstrate that learning projection matrices from data is indeed beneficial: it leads to significantly higher solution quality than the existing random projection while greatly reducing the time for solving LPs.