A team of researchers at IBM has announced an experimental system using phase-change memory that is capable of mimicking the human brain and how it processes images and video footage.
The team at IBM designed the “neuromorphic” system using phase-change memory cells which store data more densely, facilitating a greater number of synapses, and is capable of reprogramming or “learning” as it is fed new information.
Inspired by the human brain, the new system is comprised of over 900 “neurons”, or phase-change memory cells, and achieved over 165,000 synapses, where previous efforts resulted 100 synapses or less. By altering how the virtual neurons influence one another, the system has learned to recognize handwritten numbers within 82% accuracy.
