Assessing Mechanisms of Neural Synaptic Activity

Authors: 
Mike West, Guoliang Cao
Duke University

Nov 30 1990

Masses of data relating to the synaptic activity of stimulated nerve tissues is routinely generated in neurophysiological experiments designed to investigate the stochastic behavior of neurons, individually and collectively. Common experiments results in observed time series measuring evoke neural responses following various levels of electrical stimulus of the nerve tissue, and anayses typically focus on simple numerical summaries, such as estimates of maximum levels of response under given stimuli. Issues of wide interest in the neurosciences are being addressed using representations of such data using mixtures of uncertain numbers of normal distributions. Focuses of analysis are on mixture deconvolution, and on the numbers ad weights of components. One approach currently under study uses Dirichlet process mixtures of normals. Some of the scientific issues, together with technical aspects of data analysis, modeling, and the use of prior information, are described and exemplified in this paper.

Keywords: 

deconvolution of mixtures, mixtures of normal distributions, neural response machanisms

Manuscript: 

PDF icon 1991-18.pdf