Project: A self-navigating Agent capable of perception and building a 2D spatial Map of its environment.
Blog: https://decision-machines.com/blogs/f/brain-based-spatial-navigation
Project Description:
The project involves building of Hippocampal-complex, Entorhinal Complex, Thalamus & Neocortical layers. The project assumes that these 4 structures together are sufficient and necessary for the perception and persistence of a complete spatial cognitive map for a primitive Perception Agent. All of these systems are based on the core principle of Spiking HTM Neural networks which communicate with each other on the fundamental principles of Sparse Distributed Representations. I am willing to go over the entire system with anybody who is willing to learn and collaborate. I have completed a good prototype of the neocortical part of the system with sequence memory implemented very well. right now, am overwhelmed with the amount of work that still needs to be accomplished. any interested folks are most welcome to join me on the exciting endeavor.
Hippocampal Functions:
Hippocampal complex for all necessary purposes can be assumed as the “THE GREAT STITCHER” it stitches all the sensory glimpses together for a short period of time, ex: all sensory modalities like touch, vision, olfactory, sound with their exact locations and embeds all of this together into a one single memory and decides the importance of these memories for Long Term Potentiation.
Thalamus Functions:
Thalamus can be assumed to be the big “Relay Engine”, it is the big network hub that is used for filtering all the inputs and sensory information and relaying all the information for inter region communication.
Entorhinal Cortex Complex Functions:
The Entorhinal Complex is completely responsible for providing the exact coordinates for every single sensory input that is being sensed by neocortex, it could be an allocentric location or an egocentric location.
Neocortex Functions:
Neocortex is the most complex, multi-functional evolutionary marvel that has every neuroscientist in a fix, at the risk of over-simplifying it, I look at it as very complexly interconnected mesh of different types of neurons in multi layered architecture to process the above allocentric and egocentric sensory features.
Code Link: GitHub - deric18/Hentul: Screen Grabber Application based out of Hierarchical Temporal Memory