Scientists from the New Jersey Institute of Technology, working with fellow researchers at IBM Research Zurich Laboratory and the École Polytechnique Fédérale de Lausanne, have demonstrated a novel synaptic architecture which presents the possibility of creating a new class of information processing systems, based on the human brain. These types of computers have a potential use for data processing.
In recent years deep learning algorithms have been used to solve complex cognitive tasks, like controlling autonomous vehicles and for language interpretation. Such algorithms are based on artificial neural networks, which are types of mathematical models of the neurons and synapses of the brain. These networks process large amounts of data and the synaptic strengths are adjusted to learn the required patterns hidden into different data streams.
Many algorithms are relatively inefficient and they require considerable quantities of power and time to process. This has led scientists to seek new materials and systems to incorporate the algorithms. These are termed nanoscale memristive devices, which are electrical machines where the conductivity depends on prior signaling activity. These devices can function to represent the synaptic strength between the neurons in artificial neural networks.
By taking prototype chips, designed by IBM, and containing over one million nanoscale phase-change memristive devices, the researchers showed how how multiple nanoscale memristive devices can be successfully configured to effectively implement artificial intelligence algorithms for improved deep learning.
The study has been published in Nature Communications and the research paper is titled “Neuromorphic computing with multi-memristive synapses.”
