Sign up for our newsletter
Technology / AI and automation

AT&T CREATES PARALLEL “NEURAL NET” CHIP TO SOLVE ROUTING PROBLEMS

A simple version of a classic problem in mathematics has been solved by a computer chip that mimics the action of nerve cells in animals, reports Microbytes Daily. Researchers at AT&T have reported that they have developed a neural network chip which can solve the celebrated travelling salesman problem, which requires the most cost-effective route to be found between a series of towns. The only solutions are sophisticated versions of trial and error and the problem rapidly becomes impossibly complex as points and routes are added to the network – and AT&T is of course interested in it because of its application to the routing of long-distance telephone calls. A neural net computer could in theory solve even extremely complex networks very quickly; animal nervous systems are marked by a very high degree of connectivity, with each nerve linked to hundreds or thousands of other ones, and by the presence of both excitory and inhibitory impulses. AT&T researchers simulated the connectivity of a nervous system by building a crossbar switch on a chip: the grid-like switch – actually an associative memory chip with each neuron representing one bit – enables all the signals in the circuit to interact with all other signals. Amplifiers act as exciters and resistors function as inhibitors. The chip is programmed by forming resistors at the appropiate points with electron-beam lithography. By using resistors rather than transistors as the inhibitors, the circuit can be made much smaller, and AT&T claims it can get 256 neurons onto a single chip. And neural net chips are fast because they operate in parallel, tackling the whole problem at once rather than sequentially. Rather like an analogue computer, the search key is entered as the initial state of the system and the chip automatically seeks the closest possible match. Once a neural net chip has been programmed, it goes to work as soon as power is applied.

White papers from our partners


This article is from the CBROnline archive: some formatting and images may not be present.

CBR Staff Writer

CBR Online legacy content.