Network optimization problems we solve
Facility location
Select warehouse, data center, or cell tower locations to minimize total service cost. Uncapacitated and capacitated facility location as a QUBO with binary site-selection variables.
Graph partitioning
Divide a network graph into balanced partitions with minimum edge cut. Used for distributed computing workload placement, VLSI design, and community detection in social networks.
Traffic engineering
Assign network flows to paths to minimize congestion and maximize throughput. Multi-commodity flow problems on backbone telecom networks.
Minimum spanning tree variants
Degree-constrained spanning trees, Steiner trees, and capacitated minimum spanning trees for physical network infrastructure design.
Sensor / antenna placement
Place the minimum number of sensors to achieve required coverage of a geographic area. Set cover formulation as a QUBO.
Graph problems and QUBO
Many network design problems have natural QUBO formulations. MaxCut, graph coloring, clique finding, and independent set problems all map directly to Ising energy minimization. NEROX supports direct graph input — supply an adjacency matrix and NEROX constructs the QUBO automatically.
Scale
NEROX handles graphs with up to 50,000 nodes in a single GPU pass. For larger networks (backbone telecom graphs, national road networks), the Hybrid Solver decomposes the graph into subgraphs, solves each independently, and stitches the global solution.