Photoelectric hybrid intelligent data generation calculation system and method

Abstract

The invention discloses a photoelectric hybrid intelligent data generation and calculation system and method, and the system comprises an electronic compression sampling module which is used for learning the feature probability distribution of input data in an unsupervised manner, compressing the input information to a low-dimensional space, and carrying out the distributed sampling; a feature conversion module which is used for converting the compressed and sampled electronic feature signal into an optical feature signal; and an all-optical data generation module which consists of a pluralityof passive optical frequency domain modulation modules and is used for generating brand-new calculation data from the input optical characteristic signals. The system can realize intelligent data generation of light speed, so that a photoelectric hybrid system or all-optical machine learning can realize unsupervised intelligent data generation.

Description

technical field [0001] The invention relates to the technical fields of photoelectric computing and machine learning, and in particular to a photoelectric hybrid intelligent data generation computing system and method. Background technique [0002] Generative models are one of the most important classes of models in machine learning. This type of model can randomly generate observation data and can be used to directly model the data or establish a conditional probability distribution between variables. It has been able to automatically generate data such as images, texts, and sounds. [0003] At present, there has been an all-optical diffraction deep neural network, which has realized an all-optical machine learning discriminant model. The architecture optimizes the combination of multi-level space-frequency domain optical phase modulation layer and nonlinear layer similar to artificial neural network through machine learning design, and realizes functions such as handwrit...

AI Technical Summary

Problems solved by technology

[0005] However, the above-mentioned optical diffraction neural network can only implement machine learning discriminative models at present, and the work in the field of high-speed data generation using optical networks is very limited.

Unsupervised use of optical networks for automatic data generation is currently not possible and needs to be resolved

Images