Authors:
Ilan Oren
;
Raghd Abu-Sinni
and
Ramez Daniel
Affiliation:
Faculty of Bio-Medical Engineering, Technion - Israel Institute of Technology, Haifa 3200003, Israel
Keyword(s):
Subthreshold Electronic Circuits, Analog Design, Neural Network, Molecular Network, Bio-Inspired.
Abstract:
Neuromorphic engineering, inspired by principles and architecture of neuronal circuitries, enabled the design of Artificial Neural networks (ANNs) for Intelligent systems. These systems perform very complex computation tasks, yet they consume significant power. Thus, using artificial intelligence (AI) for applications where only a small power source is available is very limited. While the neuronal networks in the brain can recognize complex patterns and memorize enormous elements, molecular and protein networks can perform other complex tasks such as adaptive immunity and cell differentiation at high energy efficiency. Here, we claim that a bio-inspired computing platform mimicking molecular protein networks can lead to ultra-low power emergent computation. Previously, we proposed a molecular-inspired computing model named Perceptgene that has the attributes of learning and adaptivity as the neural network (Rizik et al., 2022). Similarities were found between equations describing bio
chemical reactions and transistor operation at subthreshold (Sarpeshkar, 2011) enabling the design of Perceptgene with subthreshold electrical circuits. Thus, the subthreshold Perceptgene circuits are expected to allow computing and learning capabilities at ultra-low power consumption.
(More)