Massachusetts Institute of Technology

Sync Computing Corp.

Jeffrey Chou

Suraj Deepak Bramhavar

Jeffrey G Bernstein

Sang-Gook Kim

William D Herzog

Erica Lin

Siddhartha Ghosh

Christopher Roberts

Titash Rakshit

Juejun Hu

Vladimir Liberman

Yifei Zhang

Jason Stewart

Christopher A D Roeser

Luna Xu

Marco Montes De Oca

A system including digital oscillators and at least one programmable interconnect is described. The programmable interconnect(s) provide weights for and selectably couples at least a portion of the digital oscillators. The digital oscillators and the programmable interconnect(s) form an optimization processing unit (OPU). A system for performing reversible logic is also described. The system includes digital oscillators coupled to perform a logic operation and an error correction unit coupled to the digital oscillators. The error correction unit is configured to sample states of the digital oscillators, detect error(s) in the states, and tune connection coefficient(s) between the oscillators in response to detecting the error(s).

Optimization processing unit utilizing digital oscillators

A method of efficiently executing a plurality of processes is described. The method generates, using a predictor, operating constraints for the processes. An operating constraint of the operating constraints is for each process of the processes. The method evaluates possible operating points for each process consistent with the operating constraints and according to a cost function for the processes. The method also determines an operating point for each process based on the evaluating.

Automated globally optimized resource allocator and scheduler

A method of assigning processing resources is described. The method includes receiving an application and analyzing the application to determine an expected run time use of the processing resources. At least a portion of the processing resources are assigned to the application based on the expected run time use of the processing resources. A computing system architecture including a processing resource pool, an interface, and a special purpose optimization coprocessor is also described. The interface receives an application to be executed. The special purpose optimization coprocessor receives information about the application and the processing resource pool and outputs an allocation scheme for allocating tasks of the application to the processing resource pool.

Real time scheduling using expected application resource usage

An analog computing system with coupled non-linear oscillators can solve complex combinatorial optimization problems using the weighted Ising model. The system is composed of a fully-connected LC oscillator network with low-cost electronic components and compatible with traditional integrated circuit technologies. Each LC oscillator, or node, in the network can be coupled to each other node in the array with a multiply and accumulate crossbar array or optical interconnects. When implemented with four nodes, the system performs with single-run ground state accuracies of 98% on randomized MAX-CUT problem sets with binary weights and 84% with five-bit weight resolutions. The four-node system can obtain solutions within five oscillator cycles with a time-to-solution that scales directly with oscillator frequency. A scaling analysis suggests that larger coupled oscillator networks may be used to solve computationally intensive problems faster and more efficiently than conventional algorithms.

All-to-all connected oscillator networks for solving combinatorial optimization problems

Techniques usable in optimization processing are described. A system includes an optimization processing unit (OPU). The OPU includes stochastic computing units and at least one programmable interconnect. Each of the stochastic computing units includes nodes and multiplication unit(s) configured to interconnect at least a portion of the nodes. The programmable interconnect(s) are configured to provide weights for and to selectably couple a portion of the stochastic computing units.

Optimization processing unit having subunits that are programmably and partially connected

Networks of superharmonic injection-locked (SHIL) electronic oscillators can be used to emulate Ising machines for solving difficult computational problems. The oscillators can be simulated or implemented in hardware (e.g., with LC oscillators) and are coupled to each other with links whose connection strengths are weighted according to the problem being solved. The oscillators' phases may be measured with respect to reference signal(s) from one or more reference oscillators, each of which emits a reference signal but does not receive input from any other oscillator. Sparsely connected networks of SHIL oscillators and reference oscillators can be used as Viterbi decoders that do not suffer from the information bottleneck between logic computational blocks and memory in digital computing systems. Sparsely connected networks of SHIL oscillators and reference oscillators can also be programmed to act as Boolean logic gates that operate in both forward and backward directions, enabling multipliers that can factor numbers.

All electrical fully connected coupled oscillator Ising machine

Alloys of GeSbSeTe (GSST) can be used to make actively tunable infrared transmission filters that are small, fast, and solid-state. These filters can be used for hyperspectral imaging, 3D LIDAR, portable bio/chem sensing systems, thermal emission control, and tunable filters. GSST is a low-loss phase-change material that can switch from a low-index (n=3), amorphous state to a high-index (n=4.5), hexagonal state with low loss (k&#x3c;0.3) over a wavelength range of 2-10 microns or more. The GSST thickness can be selected to provide pure phase modulation, pure amplitude modulation, or coupled phase and amplitude modulation. GSST can be switched thermally in an oven, optically with visible light, or electrically via Joule heating at speeds from kilohertz to Gigahertz. It operates with reversible and polarization independent transmission switching over a wide incident angle (e.g., 0-60 degrees).

Methods and apparatus for modulating light with phase change materials

A metallic-dielectric photonic crystal is formed with a periodic structure defining a plurality of resonant cavities to selectively absorb incident radiation. A metal layer is deposited on the inner surfaces of the resonant cavities and a dielectric material fills inside the resonant cavities. This photonic crystal can be used to selectively absorb broadband solar radiation and then reemit absorbed radiation in a wavelength band that matches the absorption band of a photovoltaic cell. The photonic crystal can be fabricated by patterning a sacrificial layer with a plurality of holes, into which is deposited a supporting material. Removing the rest of the sacrificial layer creates a supporting structure, on which a layer of metal is deposited to define resonant cavities. A dielectric material then fills the cavities to form the photonic crystal.

Growing community of inventors

Walnut Creek, CA, United States of America

Jeffrey Chou