Solving a general integer program (IP) is NP-hard. The classic algorithm [Papadimitriou, J.ACM '81] for IPs has a running time n^{{𝒪}(m)}(m⋅max{Δ,‖b‖_{∞}})^{{𝒪}(m²)}, where m is the number of constraints, n is the number of variables, and Δ and ‖b‖_{∞} are, respectively, the largest absolute values among the entries in the constraint matrix and the right-hand side vector of the constraint. The running time is exponential in m, and becomes pseudo-polynomial if m is a constant. In recent years, there has been extensive research on FPT (fixed parameter tractable) algorithms for the so-called n-fold IPs, which may possess a large number of constraints, but the constraint matrix satisfies a specific block structure. It is remarkable that these FPT algorithms take as parameters Δ and the number of rows and columns of some small submatrices. If Δ is not treated as a parameter, then the running time becomes pseudo-polynomial even if all the other parameters are taken as constants. This paper explores the trade-off between time and accuracy in solving an IP. We show that, for arbitrary small ε > 0, there exists an algorithm for IPs with m constraints that runs in {f(m,ε)}⋅poly(|I|) time, and returns a near-feasible solution that violates the constraints by at most εΔ. Furthermore, for n-fold IPs, we establish a similar result - our algorithm runs in time that depends on the number of rows and columns of small submatrices together with 1/ε, and returns a solution that slightly violates the constraints. Meanwhile, both solutions guarantee that their objective values are no worse than the corresponding optimal objective values satisfying the constraints. As applications, our results can be used to obtain additive approximation schemes for multidimensional knapsack as well as scheduling.

Mixed-integer linear programming (MILP) is at the core of many advanced algorithms for solving fundamental problems in combinatorial optimization. The complexity of solving MILPs directly correlates with their support size, which is the minimum number of non-zero integer variables in an optimal solution. A hallmark result by Eisenbrand and Shmonin (Oper. Res. Lett., 2006) shows that any feasible integer linear program (ILP) has a solution with support size $s\leq 2m\cdot\log(4mΔ)$, where $m$ is the number of constraints, and $Δ$ is the largest coefficient in any constraint. Our main combinatorial result are improved support size bounds for ILPs. To improve granularity, we analyze for the largest $1$-norm $A_{\max}$ of any column of the constraint matrix, instead of $Δ$. We show a support size upper bound of $s\leq m\cdot(\log(3A_{\max})+\sqrt{\log(A_{\max})})$, by deriving a new bound on the -1 branch of the Lambert $\mathcal{W}$ function. Additionally, we provide a lower bound of $m\log(A_{\max})$, proving our result asymptotically optimal. Furthermore, we give support bounds of the form $s\leq 2m\cdot\log(1.46A_{\max})$. These improve upon the previously best constants by Aliev. et. al. (SIAM J. Optim., 2018), because all our upper bounds hold equally with $A_{\max}$ replaced by $\sqrt{m}Δ$. Using our combinatorial result, we obtain the fastest known approximation schemes (EPTAS) for the fundamental scheduling problem of makespan minimization of uniformly related machines ($Q\mid\mid C_{\max}$).