A New Library for Network Optimization

Tuesday, February 22, 2022

Networks are all around us from the electrical circuits inside our computers to the multitude of internet servers that route packets of data around the globe. Even the web itself is a network of pages connected to each other by a myriad of blue links.

A network of rubber bands on a wooden pegboard

A network's structure is referred to as its topology. Network topologies can be physical or logical, centralized or decentralized, and fully or partially connected. Given a network with n nodes, the number of possible topologies grows exponentially with n; even just a dozen nodes admit nearly a trillion trillion possible configurations!

The network-opt logo

We are pleased to announce the open source release of network-opt, a C++ library that supports the optimization of network topologies. Using sophisticated techniques for combinatorial search, this algorithm can efficiently construct instances from a family of so-called series-parallel networks that commonly arise in electrical and telecommunications applications. For example, given 15 1-Ω resistors and a target resistance of π Ω, our tool can produce a circuit that achieves six digits of precision:

A 15-element circuit composed of 1-ohm resistors

For more details, refer to our paper: "Search Strategies for Topological Network Optimization," appearing this month at the Thirty-Sixth AAAI Conference on Artificial Intelligence (AAAI-22).

By Michael D. Moffitt, Core Enterprise Machine Learning