Bio-Inspired Adaptive Integrated Information Processing

Abstract

Biological brains are dramatically more effective in dealing with real-world adaptive information processes and decisions than most advanced computers. Advanced computers can utilize the discipline of classical signal processing whereby providing theoretical mathematical and statistical approaches for information processing, and with the vision of bio-inspired adaptive processing are evolving into neuromorphic integrated sensory processing systems. Numerous MISO (multi-input, single output) sensory applications demand reliable and effective information processing which, at an initial stage may be addressed by solving the key problems of advanced computing platforms, which are: i-massive parallelism; ii-low power consumption of massively large systems; iii-intelligent systems that learn from observations and perform better on the next run; iv-integrated systems for embedded feasibility; and v-systems that adapt to the environment. Thus, our Adaptive Integrated Information Processing (AIIP) approaches, presented in this paper. Two AIIP systems are presented: Neural networks with synaptic plasticity, as our Spiking Neural Networks (SNNs), with up to one million inputs, for chemical sensing and detection; and Adaptive Recurrent Dynamic Neural Networks (ARDNNs) for defect tracking, reliable system deployment, and prognosis for telecommunication and similar applications. Further-our presented AIIP systems may provide a viable solution to offering powerful modulation schemes and transmission rates far beyond current possible communications systems.