A novel compression based neuronal architecture for memory encoding

Research in neuro-biological memory encoding suggests that it takes place through various biophysical and biochemical mechanisms during synaptic transmission of information between neurons. However, there are no mathematical models to explain how these processes result in real-time memory encoding which is compressed and distributed in different neuronal pathways across different brain regions. Biologically inspired artificial neural networks that accomplish learning by updating its synaptic weights, lack a theoretical justification. In this work, we propose a novel biologically inspired network architecture of neural memory encoding, preserving its various attributes including compression, non-linearity, distributed processing and dynamical nature. We demonstrate that our model is capable of universal computation and satisfies the approximation theorem.