PI Institution: Missouri State University
The Goal of this Research:
The goal of this project is to research and develop a new kind of small, economical, optical neural computer which will be ten times as fast as today's fastest supercomputers. It will have the ability to rapidly learn new tasks without requiring costly and complex programming. The approach is to construct neural networks which can dynamically learn tasks by encoding the learned information in recurrent optical signal loops. This method is similar to short-term memory in humans.
Intellectual Merit:
The intellectual merit of this project is increased understanding of the software and hardware for Fixed-Weight Learning and Optical Neural Networks, the scientific areas on which this project is based. It will move this kind of neural network from a theoretical concept to a prototype capable of performing real-world tasks.
Broader Impacts:
Small optical neural computers that are ten times as fast as the fastest supercomputer would have wide ranging applications across all of society. A few examples are advanced robots used in areas such as Disaster Response and Unmanned Air Vehicles, smarter medical devices, and improved speech and image recognition. The neural network's ability to rapidly learn would allow the development of systems that can adapt to changing environments without costly and time-consuming reprogramming.
It will also give undergraduate students at Missouri State University the opportunity to learn valuable investigative skills and gain practical experience in an advanced research and development project.
