Theory of Glial Cells &Neurons Emulating Biological Neural Networks (BNN) for Natural Intelligence (NI) Operated Effortlessly at A Minimum Free Energy (MFE)
Harold H Szu · MOJ Applied Bionics and Biomechanics · 2017
Recent studies in neurosciences have produced evidence suggesting that Glial cells play a vital role in assisting neurons to form synaptic connections, contrary to a previously held view that they played a less significant role in the synaptic function.The neurochemical processes underlying the transmission is observed to be both temperature dependent, and ionic conduction, enabling faster communication.These results have opened up a few key questions: what role do Glia play in information processing, learning, memory, and retention in the brain?This opens up the possibility for on-demand interconnection for low-latency synaptic connection between otherwise unconnected nodes, and their importance to scale the performance of very large neural networks, typically trained by deep learning, but designed to act on inputs originating from very large geo-distributed areas.In this paper, we propose a new framework for Unsupervised Deep Learning (UDL) that takes into account the existence of such interconnections.The theory is based on the thermodynamic equilibrium of human brains that are kept at a constant temperature to make effortless decisions of all incoming sensory data at the Helmholtz Minimum Free Energy (MFE).This is referred to as a Natural Intelligence (NI) as opposed to AI in the sense of a noncontrived straightforward decision.Likewise, the trustworthiness of MFE classifier will be comprehensible with Human Visual System (HVS) with sparse Ortho-Normal (ON) Salient Feature Extraction (SFE).The MFE cost function is derived from first principles obtained from nature: one, the Homeostasis principle; and two, real-time duplicative sensory inputs.The Homeostasis condition maintains constant brain temperature, which implies constant biochemistry reaction rates resulting in the same learning experience among all generations of Homosapiens.The Power of Paired sensory inputs from eyes, ears, nostrils, tessellate tasting buds, tactile touching sensing has real-time pre-processing that exploits the agreement is the signal, while the disagreements are the noises, and the input signal energy relaxes to the averaged brain temperature as the UDL.A mathematical definition of Glial (Greek: glue) cells seems to corroborate with modern neuroscience of living animals.There are tens of billions of Neurons and hundreds of billions of Glial cells that keep our brains operating smoothly.Any imbalance between two can generate disorders.Thus the crossover Rosette Stone between ANN Back Propagation delta vs. glial and dendrites vs nets as well as BNN UDL ANN SDL.These cross-over will be facilitated for medical image screening and tracking.While a living system might be sick by either DNA genome or epigenetic phoneme, the machine can likewise crash by programming bugs.For example, when the glial force be too strong divergent, implicating the Glioma tumors, when Astrocytes glial servant cells go on strike: slow down (lacking of deep sleep) leading to Alzheimers.We believe the artificial machine can compute the cause and the effect helping radiologists' screenings and diagnoses.The experimental characterization of those electrical insulated white matter made of six types of Glial cells in live animals might help proactively & early diagnose by means of powerful machine learning, and thus improve treatment of the disorders of human nervous system.