FemtoComputing: machine learning using femtosecond pulses and nonlinear optics
We report a new concept in hardware acceleration of AI that exploits femtosecond pulses for both data acquisition and computing. Data is first modulated onto the spectrum of a supercontinuum laser. Nonlinear optical propagation then projects the data into an intermediate space in which data classification accuracy is enhanced. This nonlinear optical kernel operation improves the linear classification results similar to a traditional numerical kernel (such as the radial-basis-function) but with orders of magnitude lower latency. The performance is data-dependent due to the limited degrees of freedom in the optical part of the system.
UCLA Samueli School of Engineering (United States)