Skip to main content

Library Item

Signal Propagation Through Neural Networks


Researchers have attempted to describe the propagation of information through neural networks using theoretical models. A recent model has shown that it is theoretically possible for a small neural network to compute a vast range of functions required to do many complex tasks. In these models, a task-dependent network design is not required due to the inherent, high-dimensional dynamics in neural networks. To test this theory in a biological network, I inject a sine wave of current into a cell within the stomatogastric ganglion (STG) of the crab and measure the firing patterns of the remaining STG cells in response to this injected stimulus. Due to the complexity of the network, the sine wave of current is “translated” into firing patterns that encode the features of the sine wave in terms of complex signals. These data can be analyzed in many different ways. Spectral analysis of these data will show the power of the injected frequency in each cell in the network, thereby providing an estimate of the linear range of the signal. Principal components analysis will set the foundation for probing the dynamic range of the network in terms of the nonlinear transformations of the input.