Steady-state Learning and Synaptic Connectivity in Local Cortical Networks of Excitatory and Ihibitory Neurons
Author | : Julio Ivan Chapeton |
Publisher | : |
Total Pages | : 90 |
Release | : 2014 |
Genre | : Associative storage |
ISBN | : |
Download Steady-state Learning and Synaptic Connectivity in Local Cortical Networks of Excitatory and Ihibitory Neurons Book in PDF, ePub and Kindle
Learning and memory storage are arguably the most fundamental and well-studied functions of the mammalian cortex. It is established that these functions are mediated by many forms of synaptic plasticity, which shape neural circuits in the course of learning by creating, modifying, and eliminating individual synaptic connections. Nevertheless, the effects of learning and memory storage on the cortical connectivity diagram in the adult are largely unknown. In general, it is difficult to find examples where the link between a function and the connectivity of the underlying neural circuit is completely understood. Experiments have shown that some connectivity features are ubiquitously present in local cortical networks. These features include very sparse connectivity of excitatory neuron axons, much denser connectivity established by the axons of many inhibitory neuron classes, and stereotypically distributed connection weights. Given the pervasiveness of these features, is it possible that they could have arisen as a direct consequence of learning? To answer this question, in Chapter 2 we examine a biologically realistic, yet exactly solvable model of associative memory which is based on the hypothesis that synaptic connectivity in a given local circuit of adult cortex is in a steady-state; in this state the associative memory storage capacity of the circuit is maximal and learning of new associations is accompanied by forgetting of some of the old ones. The model is applicable to networks of multiple excitatory and inhibitory neuron classes and can account for homeostatic constraints on the number and the overall weight of functional connections received by each neuron. In Chapter 3 we describe how the model was solved analytically by using the replica theory from statistical physics, and we highlight the most salient features of synaptic connectivity which arise from steady-state learning. Chapter 4 is devoted to testing the validity of the model by comparing these features with a large dataset of published experimental studies reporting amplitudes of unitary postsynaptic potentials and probabilities of connections between various classes of excitatory and inhibitory neurons in the cerebellum, neocortex, and hippocampus. The theoretical results are in good agreement with these experimental measurements, suggesting that stereotypic features of adult connectivity can form despite functional differences among brain areas and diverse learning experiences of individual animals. Lastly, in Chapter 5 we show how biologically constrained learning can be used in a machine learning methodology to accurately trace sparsely labeled neurites in light microscopy stacks of images.