Page 16

EETE FEB 2016

ENSW & CETHNOLOGY neural networks MIT neural network IC aims at mobiles By Peter Clarke MIT researchers have designed a chip to implement neural networks and claim it is 10 times as efficient as a mobile GPU. This could be used to allow mobile devices to run artificial-intelligence algorithms locally, rather than uploading data to the Internet for processing. MIT researchers have presented a chip designed specifically to implement neural networks at the International Solid State Circuits Conference in San Francisco. “Deep learning is useful for many applications, such as object recognition, speech, face detection,” said Vivienne Sze, a professor in MIT’s Department of Electrical Engineering and Computer Science whose group developed the chip. The chip is called Eyeriss and the idea is that training up the network for specific functions could be done “in the cloud” and then the training configuration – the weighting of each node in the network could be exported to the mobile device. The chip has 168 cores each with its own memory with the idea to minimize the frequency with which cores need to exchange data with distant memory banks. When such exchanges are off-chip in particular it consumes much time and energy. The chip has a circuit that compresses data before sending it to individual cores and each core is also able to communicate directly with its immediate neighbors, so that if they need to share data, they don’t have to route it through main memory. The final key to the chip’s efficiency is a circuit that allocates tasks across cores. In its local memory, a core needs to store not only the data manipulated by the nodes it’s simulating but also data describing the nodes themselves. The allocation circuit can be reconfigured for different types of networks and applied across the network at the start of running the application. As part of the ISSCC presentation the MIT researchers used Eyeriss to implement a neural network that performs imagerecognition. “In addition to hardware considerations, the MIT paper also carefully considers how to make the embedded core useful to application developers by supporting industry-standard network architectures AlexNet and Caffe,” said Mike Polley, a senior vice president at Samsung’s Mobile Processor Innovations Lab. The MIT researchers’ work was funded in part by DARPA. Neural network built in plastic By Peter Clarke A team of scientists from Russia and Italy have created a neural network made from plastic memristors. Kurchatov Institute, MIPT, the University of Parma, Moscow State University, and St. Petersburg State University have demonstrated that it is possible to create simple polyaniline-based neural networks that are able to learn and perform logical operations. The work is reported in Organic Electronics. A memristor is an electrical element with a variable resistance that depends on the charge passing through it and which displays memory. It has long-been noted that a memristor is similar to a biological synapse – a connection between two neurons in the brain that is able to modify the efficiency of signal transmission between neurons under the influence of the transmission itself. To date studies of synapse-like behavior have been done in silicon-based memristor-type devices. Using a polyaniline solution, a glass substrate, and chromium electrodes, team created a polymeric memristor at a scale of about 1 millimeter and multiple memristors were then connected in a single neuromorphic network. Under supervised learning the memristive network is capable of performing NAND or NOR logical operations. This perceptron node is of macroscopic dimensions and with a characteristic reaction time of tenths or hundredths of a second, but the researchers state that this should scale with dimensions and the neural network is made using inexpensive materials. In addition, polyaniline can be used in attempts to make a three-dimensional structure by placing the memristors on top of one another in a multitiered structure. Typical applications for neuromorphic computers include machine vision, hearing, and emulating other sensory organs, and also intelligent control systems for various devices, including autonomous robots and drones. 16 Electronic Engineering Times Europe February 2016 www.electronics-eetimes.com


EETE FEB 2016
To see the actual publication please follow the link above