Although “neural network models” have guided the evolution of software and algorithms mimicking the brain, the human brain can still perform many tasks that current computer chip technology cannot.
Karlheinz Meier, a physicist at the University of Heidelberg, says that neuromorphic engineers aim at supporting three characteristics that brains have and computers do not: low power consumption (human brains use little wattage against computers); fault tolerance (losing just one transistor can wreck a microprocessor, but brains lose neurons all the time); and a lack of need to be programmed (brains learn and change spontaneously as they interact with the world).
Considering these three characteristics, two fundamentally different approaches to the neuromorphic chips technology can be observed:
1. The above three characteristics can be achieved through manufacturing a fully hardware architecture mimicking the brain and running without the help of cloud computing.
2. Business drives the development of a solution based
…show more content…
A node sits on a circuit and measures electrical pulses transmitted along. If a certain number of spikes occur within a certain period of time, the node is programmed to send along its own new spikes which are weighted to a range of values. Because neurons are only activated when they receive a spiking signal, the chips save on energy. Spiking neurons are programmed into complex networks. For example, a network designed for image recognition will associate the weighted connections and spiking neurons with different objects. One pattern of spikes appearing at the output will have been programmed as the image of a cat; another pattern of spikes will indicate the image is of a chair. The neuromorphic chip’s own evaluation of similarities to a programmed image or sound or signal or situation is what defines it