Towards digitizing the hippocampus a bio inspired artificial neural network

Abstract

The human brain is an amazing piece of machinery performing various tasks with varying levels of complexity. The underlying powerhouse of the human brain is the neural network it contains. Learning and memory are the two key aspects of the brain, which bring together all the sensory information from the past and the present. This enables the brain to revisit these memories that eventually helps in the process of decision making. So, how does the brain store and retrieve memories? The answer to this lies in the hippocampus region of the brain. The hippocampus is responsible for the storage and retrieval of memories in the brain. These memories are represented as patterns of neuronal activity. The type of memory handled by the hippocampus is the declarative memory, which is of two types. The first being semantic memory, that is the memory of facts and concepts. The second type is called episodic memory, it is the memory of events. This work develops an Artificial Neural Network (ANN), inspired by the working of the hippocampus region of the brain, which can handle semantic memory and episodic memory. The first challenge is to survey the vast neuroscience literature, in order to pin down the functionalities of the hippocampus. This is followed by a survey of the neural networks literature, in order to identify the gap in literature pertaining to hippocampal modeling. The next challenge is to build an ANN, which follows a complementary mid-line between the functionality of the neural network in our brain and the current methodologies in the field of Computer Science. The network proposed in this work aims to use the topology and the working of the hippocampus region, as leverage, to build an ANN. It does not try to model the hippocampus at a neuron to neuron level. newline