Reconstruction of Brain Networks by Algorithmic Amplification of Morphometry Data
IWANN '99 Proceedings of the International Work-Conference on Artificial and Natural Neural Networks: Foundations and Tools for Neural Modeling
Polychronization: Computation with Spikes
Neural Computation
Analytical Model for the Effects of Learning on Spike Count Distributions
Neural Computation
Gamma oscillations in a minimal CA3 model
Neurocomputing
| Hi-index | 0.00 |
The mammalian hippocampus is involved in spatial representation and memory storage and retrieval, and much research is ongoing to elucidate the cellular and system-level mechanisms underlying these cognitive functions. Modeling may be useful to link network-level activity patterns to the relevant features of hippocampal anatomy and electrophysiology. Investigating the effects of circuit connectivity requires simulations of a number of neurons close to real scale. To this end, we construct a model of the hippocampus with 16 distinct neuronal classes (including both local and projection cells) and 200,000 individual neurons. The number of neurons in each class and their interconnectivity are drawn from rat anatomy. Here we analyze the emergent network activity and how it is affected by reducing either the size or the connectivity diversity of the model. When the model is run with a simple variation of the McCulloch-Pitts formalism, self-sustaining non-repetitive activity patterns consistently emerge. Specific firing threshold values are narrowly constrained for each cell class upon multiple runs with different stochastic wiring and initial conditions, yet these values do not directly affect network stability. Analysis of the model at different network sizes demonstrates that a scale reduction of one order of magnitude drastically alters network dynamics, including the variability of the output range, the distribution of firing frequencies, and the duration of self-sustained activity. Moreover, comparing the model to a control condition with an equivalent number of (excitatory/inhibitory balanced) synapses, but removing all class-specific information (i.e. collapsing the network to homogeneous random connectivity) has surprisingly similar effects to downsizing the total number of neurons. The reduced-scale model is also compared directly with integrate-and-fire simulations, which capture considerably more physiological detail at the single-cell level, but still fail to reproduce the full behavioral complexity of the large-scale model. Thus network size, cell class diversity, and connectivity details may all be critical to generate self-sustained non-repetitive activity patterns.