What will it take to further increase the broad practical utility of computation?

Optimization problems lie at the heart of transportation and logistics, utilities management and climate change, public policy, engineering design, pharmaceuticals, machine learning, artificial intelligence, and computer vision. In these contexts, optimization is embedded in heterogeneous workflows that foreground energy constraints, price-performance tradeoffs, uncertainty, transparency, human factors, and cloud computing.

The NSF Expedition in Coherent Ising Machines is a 5-year, $10M project focusing on the development and assessment of new classes of specialized computing architectures for optimization. The project team spans mathematical analysis of optimization, nanophotonic and quantum optical technologies, simulation and benchmarking of unconventional computing architectures, and practical applications of machine learning.

The team was brought together by complementary engagement with testing and analysis of Coherent Ising Machines, and is pursuing an expansive agenda in physics-based optimization.

The CIM architecture is not simply a neural network or an annealer. It is inspired by those predecessors, but uses driven-dissipative optical dynamics to encode and search Ising Hamiltonians.

A conceptually straightforward CIM can be constructed as an optically pumped network of coupled degenerate optical parametric oscillators. Couplings among the oscillators encode a specific Ising Hamiltonian, and bifurcations in the network dynamics guide the search process.