A group of scientists from Northwestern University, Boston College and Massachusetts Institute of Technology has announced the successful creation and testing of a new synaptic transistor who can integrate into neural networks for associative training. This development solves the problem of the absence of a corresponding memory cell in traditional computers that constantly transmit data between the processor and memory banks.
The main advantage of a new transistor is its ability to operate at room temperature with minimal energy consumption, which is only 20 PVT (pikovatt). The researchers used the concept of moraine quantum materials, where one layer consists of graphene and the other - of boron nitride. The location of the layers provides a mora pattern, which is a key element for identifying magical angles where expressive interactions occur.
Scientists went further, creating neural chains based on these transistors who revealed the ability to associative learning. Experiments have shown that neural networks are able to recognize and separate certain combinations in binary coding. This development opens new prospects for effective memory computing schemes and involves the use in advanced hardware accelerators for artificial intelligence and machine learning.