Method and system for machine learning using optical data

A method, comprising:

• determining, using a computer processor, first error data of an electronic model using training data and predicted data that is generated by the electronic model;
• obtaining, by the computer processor and from an optical detector, optical data regarding a first combined optical signal comprising a reference optical signal and a resulting optical signal,
• wherein the resulting optical signal is produced by transmitting an optical signal through a medium at a predetermined light modulation based on the first error data;
• determining, using the computer processor, a first plurality of synthetic gradients based on the optical data, the electronic model, and the training data;
• and
• updating, using the computer processor, the electronic model using the first plurality of synthetic gradients.

A non-transitory computer readable medium storing instructions executable by a computer processor, the instructions comprising functionality for:

• determining error data of an electronic model using training data and predicted data that is generated by the electronic model;
• obtaining, from an optical detector, optical data regarding a combined optical signal comprising a reference optical signal and a resulting optical signal,
• wherein the resulting optical signal is produced by transmitting an optical signal through a medium at a predetermined light modulation based on the error data;
• determining a plurality of synthetic gradients based on the optical data;
• and
• updating the electronic model using the plurality of synthetic gradients.