The Ozcan Lab at UCLA has created optical neural networks using 3D printing and lithography. TensorFlow models were trained on the MNIST, Fashion-MNIST, and CIFAR-10 data sets using beefy GPUs. The trained models were then translated into multiple diffractive layers. These layers create the optical neural network. What the model lacks in adaptability it gains in speed as it can make predictions “at the speed of light” without any power. The basic workflow involves passing light through an input object which is filtered through the entire optical neural network to a detector which captures the results.
…each network is physically fabricated, using for example 3-D printing or lithography, to engineer the trained network model into matter. This 3-D structure of engineered matter is composed of transmissive and/or reflective surfaces that altogether perform machine learning tasks through light-matter interaction and optical diffraction, at the speed of light, and without the need for any power, except for the light that illuminates the input object. This is especially significant for recognizing target objects much faster and with significantly less power compared to standard computer based machine learning systems, and might provide major advantages for autonomous vehicles and various defense related applications, among others.
If you’d like to learn more about Photonics checkout the research happening at the Ozcan Lab. If you’d like more details about diffractive deep neural networks checkout this publication in Science or the most recent Ozcan Lab publication on the topic.
62 Replies to “Optical Machine Learning with Diffractive Deep Neural Networks #MachineLearning #3Dprinting #DeepLearning #NeuralNetworks #TensorFlow @InnovateUCLA”
Comments are closed.