Explore MemComputing's track record in advanced computing, from demonstrated optimization technologies to the next generation of processors in development.
Previous products and working demonstrations, with the documents that record them.
A software coprocessor for optimization problems at scale.
The MemCPU XPC platform delivered MemComputing's architecture as a cloud service, solving large scheduling, routing and allocation problems in real time. It was the vehicle for the applications and research results below.
Demonstrated for
Work delivered on the MEMCPU platform, documented in the case studies.
Aircraft, satellite and drone-swarm scheduling for defense programs.
Under AFWERX, Space Force and Defense Innovation Unit contracts, the platform scheduled aircraft, crews and cargo, planned LEO satellite coverage for tracking moving targets, and coordinated drone swarms. It returned near-optimal plans in minutes on problems where existing solvers took hours.
Helicopter, vessel and cargo scheduling for oil and gas operations.
Selected for Chevron Technology Ventures' Catalyst program and working with BP, MemComputing applied the platform to offshore helicopter rotations, maritime delivery routing and aircraft cargo loading, each a problem where schedule quality drives operating cost.
Published results from the same architecture.
First feasible solutions to a MIPLIB benchmark open since 2010.
The MemCPU coprocessor found feasible solutions to the f2000 problem within minutes, and showed linear scaling on mining and scheduling benchmarks where commercial solvers scaled exponentially.
Higher accuracy in fewer epochs than ADAM and SGD on a particle-detection benchmark.
A MemComputing-based training method reached better accuracy in fewer epochs than ADAM and SGD on a high-energy particle detection problem, without dropout, momentum or weight decay. A companion study applied the approach to Restricted Boltzmann Machine training on MNIST.
Selected technologies in development.
AI acceleration for real-world applications.
Proprietary AI software and processor technology being developed for edge, automotive, industrial and enterprise applications, designed to run capable models within tight memory and power limits.
A processor for large-scale optimization.
Designed to bring the platform's approach to optimization into silicon, targeting real-time results on scheduling and allocation problems, with no cryogenics or exotic materials.
Image and signal processing at the edge.
Being developed for onboard analytics on platforms across space, air, land and sea, targeting environments where power and cooling run thin.
Signal processing for advanced wireless.
Designed for beamforming in advanced 5G and Massive MIMO systems, targeting a fraction of the power of conventional approaches.
