Issue link: https://iconnect007.uberflip.com/i/1327102
JANUARY 2021 I DESIGN007 MAGAZINE 83 In the digital age, data traffic is growing at an exponential rate. The demands on computing power for applications in artificial intelligence such as pattern and speech recognition in particular, or for self-driving vehi- cles, often exceeds the capacities of conventional com- puter processors. Working together with an international team, researchers at the University of Münster are devel- oping new approaches and process architectures which can cope with these tasks extremely efficiently. They have now shown that so-called photonic processors, with which data is processed by means of light, can process information much more rapidly and in parallel - something electronic chips are incapable of doing. The results have been published in the journal "Nature." Light-based processors for speeding up tasks in the field of machine learning enable complex mathematical tasks to be processed at enor- mously fast speeds (10 12 -10 15 operations per sec- ond). Conventional chips such as graphic cards or specialized hardware like Google's TPU (Tensor Pro- cessing Unit) are based on electronic data transfer and are much slower. The team of researchers led by Prof. Wolfram Pernice from the Institute of Physics and the Center for Soft Nanoscience at the University of Münster implemented a hardware accel- erator for so-called matrix multiplications, which repre- sent the main processing load in the computation of neural networks. Neural networks are a series of algorithms which simulate the human brain. This is helpful, for example, for classifying objects in images and for speech recognition. In the experiment the physicists used a so-called con- volutional neural network for the recognition of handwrit- ten numbers. These networks are a concept in the field of machine learning inspired by biological processes. They are used primarily in the processing of image or audio data, as they currently achieve the highest accuracies of classification. The results have a wide range of applica- tions. (Source: University of Münster) Bob Tise Matt Stevenson mions (particles with half-integer spin) cannot occupy the same quantum state simultaneously. In short, you are going to smash something. 6. Data sheets are helpful, but not a replacement for experience. When it comes down to it, the data sheet is considered king, but that doesn't mean you shouldn't question it when necessary. Many problems with fit can be unintuitive or difficult to diagnose, so understanding size require- ments will prevent a lot of headaches. Every dimension needs to be looked at, from pin and through-hole size, to capacitor height. Frequently these issues will crop up when replacement components are used. Watch those tolerance ranges carefully. When you are evalu- ating replacements, make sure your replacement parts fall between maximum and minimum measurements stated in your original part data sheet. You could get away with parts that are smaller in some regards, but this should be care- fully evaluated. And don't forget to pay atten- tion to pinouts and alternate packages. DESIGN007 Bob Tise is an engineer and Matt Stevenson is the VP of sales and marketing at Sunstone Circuits. To read past columns or con- tact Tise and Stevenson, click here. Light-Carrying Chips Advance Machine Learning