Searcharxiv⌕ Search

arXiv subjects

Shermin Khosravi

Publications and source records attributed to Shermin Khosravi.

2 recordsLinked to original sources

Efficient Conversion of Algorithms to Linear Programs

The expressiveness of Linear Programming (LP) has motivated research on an LP compiler that generates polynomial-size LPs for polynomial-time solvable problems, including those having exponential extension complexity. By automating modeling from high-level algorithmic descriptions, the compiler offers a promising alternative to algebraic modelling languages. However, the resulting LPs are extremely large, posing challenges for existing solvers. Recently, we have introduced two LP compiler optimizations, namely Hierarachical Synchronization Barriers (HSB) and Constraint Sharing (CS), that address this challenge by exploiting structural and similarity-based patterns derived from the source code and the design of compiler-generated LPs, beyond what conventional compiler and presolve techniques can achieve in isolation. In this paper, we provide a comprehensive evaluation of these optimization methods on fundamental combinatorial algorithms and introduce the Execution Domain Aware (EDA) optimization framework to address a limitation revealed by this analysis. The framework reinterprets structural synchronizations as a time abstraction layer within which other optimizations can target localized reductions while remaining agnostic to time-step transitions. Across a broad suite of benchmarks, HSB and CS consistently generate smaller LPs, achieving LP file-size reductions of up to 96% and 53%, respectively, while also improving numerical conditioning, solver performance, and solution quality. HSB provides the strongest and most robust overall gains. The EDA framework further broadens the range of benchmarks that can benefit from HSB when its underlying structural assumptions are less effective. We view this work as a step toward systematic generation of Compact Integer Programming formulations for problems with exponential-size Integer Programs having polynomial-time separation oracles.

cs.MS↗

Efficient Compilation of Algorithms into Compact Linear Programs

Linear Programming (LP) is widely applied in industry and is a key component of various other mathematical problem-solving techniques. Recent work introduced an LP compiler translating polynomial-time, polynomial-space algorithms into polynomial-size LPs using intuitive high-level programming languages, offering a promising alternative to manually specifying each set of constraints through Algebraic Modeling Languages (AMLs). However, the resulting LPs, while polynomial in size, are often extremely large, posing challenges for existing LP solvers. In this paper, we propose a novel approach for generating substantially smaller LPs from algorithms. Our goal is to establish minimum-size compact LP formulations for problems in P having natural formulations with exponential extension complexities. Our broader vision is to enable the systematic generation of Compact Integer Programming (CIP) formulations for problems with exponential-size IPs having polynomial-time separation oracles. To this end, we introduce a hierarchical linear pipelining technique that decomposes nested program structures into synchronized regions with well-defined execution transitions -- functions of compile-time parameters. This decomposition allows us to localize LP constraints and variables within each region, significantly reducing LP size without the loss of generality, ensuring the resulting LP remains valid for all inputs of size $n$. We demonstrate the effectiveness of our method on two benchmark problems -- the makespan problem, which has exponential extension complexity, and the weighted minimum spanning tree problem -- both of which have exponential-size natural LPs. Our results show up to a $25$-fold reduction in LP size and substantial improvements in solver performance across both commercial and non-commercial LP solvers.

cs.PL↗