Misalignment-resilient diffractive optical neural networks

A diffractive optical neural network comprising one or more layers that are resilient to misalignments, fabrication-related errors, detector noise, and/or other sources of error comprising:

• one or more optically transmissive layers arranged in an optical path, each of the one or more optically transmissive layers comprising a plurality of physical features formed on or within the one or more optically transmissive layers and having different complex-valued transmission coefficients as a function of lateral coordinates across each layer, wherein the one or more optically transmissive layers and the plurality of physical features thereon collectively define a trained mapping function between an input optical image or input optical signal to the one or more optically transmissive layers and an output optical image or output optical signal created by optical diffraction through the one or more optically transmissive layers, the trained mapping function being resilient to one or more of:
• misalignment of one or more of the optically transmissive layers, misalignment of an input plane, misalignment of an output plane, fabrication-related errors in the optically transmissive layers and/or in the diffractive network, detector noise, and/or other sources of error;
• and
• one or more optical detectors configured to capture the output optical image or output optical signal resulting from the one or more optically transmissive layers.

A diffractive optical neural network comprising one or more layers that are resilient to misalignments, fabrication-related errors, detector noise, and/or other sources of error comprising:

• one or more optically reflective layers arranged along an optical path, each of the one or more optically reflective layers comprising a plurality of physical features formed on or within the one or more optically reflective layers, wherein the one or more optically reflective layers and the plurality of physical features collectively define a trained mapping function between an input optical image or input optical signal to the one or more optically reflective layers and an output optical image or output optical signal from the one or more optically reflective layers, the trained mapping function being resilient to one or more of:
• misalignment of one or more of the optically reflective layers, misalignment of an input plane, misalignment of an output plane, fabrication-related errors in the optically reflective layers and/or the diffractive network, detector noise, and/or other sources of error;
• and
• one or more optical detectors configured to capture the output optical image or output optical signal from the one or more optically reflective layers.