An Ad-Hoc Connectedness Model for Neural Networks
The brain and its vast network of neurons is primarily considered purely from the perspective of how action potentials and neurotransmitters propagate signals. However, this view is limiting as it does not account for the impact of the electric fields that oscillate throughout the brain. These fields have traditionally been ignored because the fields from individual neurons are small and considered negligible. However, groups of neurons form clusters and their aggregate field effects are not negligible, just as a group of antennas in a phased array have significantly more transmission power than any one antenna in the array. In phased antenna arrays, the synchrony of the array elements is critical to combining the fields from the component antennas to create and steer a beam. In the neural world, the macro field oscillations due to the synchronicity of neural networks has been of great research interest, but models that explain the underpinnings of this synchronization are lacking. This work proposes a two-part model for explaining the establishment and maintenance of synchronicity in neural networks (the Ad-Hoc Connectedness Model for Neural Networks), based on unifying the results from prior research. First, plastic neural networks are greedily optimizing decentralized (ad-hoc) networks that dynamically adjust synaptic links based on a perceived usefulness for the respective links. Second, the handshaking signals used to encode the perceived utility for links are sensitive to even weak electric fields, and therefore the electric fields can modulate these greedily optimized networks.
Talavage, Purdue University.
Neurosciences|Biomedical engineering|Electrical engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our