A method for designing neural networks optimally suited for certain tasks

MIT News  March 30, 2023 While neural networks are used for classification tasks across domains, a long-standing open problem in machine learning is determining whether neural networks trained using standard procedures are consistent for classification, i.e., whether such models minimize the probability of misclassification for arbitrary data distributions. A team of researchers in the US (MIT, UC San Diego) identified, constructed and analyzed infinitely wide networks that were also infinitely deep. Using the recent connection between infinitely wide neural networks and neural tangent kernels, they provided explicit activation functions that could be used to construct networks that achieve consistency. They […]

Study urges caution when comparing neural networks to the brain

MIT News  November 2, 2022 The central claims of recent deep learning-based models of brain circuits are that they make novel predictions about neural phenomena or shed light on the fundamental functions being optimized. Through the case-study of grid cells in the entorhinal-hippocampal circuit, a team of researchers in the US (Stanford University, MIT) showed that one often gets neither. They reviewed the principles of grid cell mechanism and function obtained from analytical and first-principles modeling efforts and examined the claims of deep learning models of grid cells. Using large-scale hyperparameter sweeps and theory-driven experimentation, they demonstrated that the results […]

World’s fastest optical neuromorphic processor

Science Daily  January 7, 2021 Convolutional neural networks have been central to the artificial intelligence revolution, but existing silicon technology increasingly presents a bottleneck in processing speed and energy efficiency. An international team of researchers (Australia, China, Canada) has demonstrated an optical neuromorphic processor for artificial intelligence which operates faster than 10 trillion operations per second (TeraOPs/s) and is capable of processing ultra-large scale data, enough to achieve full facial image recognition, something that other optical processors have been unable to accomplish. The system uses a single processor and was achieved using a new technique of simultaneously interleaving the data […]

Engineers design a device that operates like a brain synapse

Nanowerk  June 19, 2020 Learning that takes place in the brain is based on the gradual strengthening or weakening of the connections between synapses whose electronic conductance can be controlled electrically. A team of researchers in the US (MIT, Brookhaven National Laboratory) has developed a resistive switch which is an electrochemical device, made of tungsten trioxide (WO3) and works in a way similar to the charging and discharging of batteries. Just as conductivity can be changed in silicon by doping, they changed the resistance of the synapses by moving ions. Calcium, potassium, magnesium ions were used. They have demonstrated good […]

All-optical diffractive neural network closes performance gap with electronic neural networks

Science Daily  August 13, 2019 Optical computing provides unique opportunities in terms of parallelization, scalability, power efficiency, and computational speed and has attracted major interest for machine learning. Researchers at UCLA have demonstrated systematic improvements in diffractive optical neural networks, based on a differential measurement technique that mitigates the strict nonnegativity constraint of light intensity. Using this differential detection scheme, involving 10 photodetector pairs behind 5 diffractive layers with a total of 0.2 million neurons, they numerically achieved blind testing accuracies of 98.54%, 90.54%, and 48.51% for MNIST, Fashion-MNIST, and grayscale CIFAR-10 datasets, respectively. By utilizing the inherent parallelization capability […]

Physicists train the oscillatory neural network to recognize images

Phys.org  March 22, 2019 An oscillatory neural network is a complex interlacing of interacting elements that can receive and transmit oscillations of a certain frequency. Based on coupled oscillator networks implemented on vanadium dioxide structures, researchers in Russia have developed a synchronization registration method with high sensitivity and selectivity. They trained the network to synchronize only for a specific input image. In the study, the input images were transmitted to the network by changing the supply currents which changed the oscillation frequencies of oscillators. As a result, the network reacted to each received image with specific dynamics. According to the […]

Machine learning technique reconstructs images passing through a multimode fiber

Science Daily  August 9, 2018 Researchers in Switzerland used deep neural networks (DNNs) to classify and reconstruct the input images from the intensity of the speckle patterns that result after the inputs are propagated through multimode fiber. They demonstrated this result for fibers up to 1 km long by training the DNNs with a database of 16,000 handwritten digits. Better recognition accuracy was obtained when the DNNs were trained to first reconstruct the input and then classify based on the recovered image. They reported remarkable robustness against environmental instabilities and tolerance to deviations of the input pattern from the patterns […]