On Integer Programming, Discrepancy, and Convolution

Integer programs with a fixed number of constraints are solvable in pseudo-polynomial time the largest coefficient any constraint. We give new algorithm which improves running state art. Moreover, we show that improving on our for is equivalent to over quadratic (min, +)-convolution. This strong evidence algorithm’s best possible. also present specialized testing feasibility an integer program and tight lower bound, based exponential hypothesis this case. Funding: work was supported by German Research Foundation [Grants JA 612/16-1, 612/20-1].