| Poster PS25A | |
|---|
| Daniel Dougherty |
|---|
| Michigan State University |
|---|
| Title | Wedge: A Bayesian Smoothing Algorithm For Neuronal Models |
|---|
| Abstract | A robust Bayesian smoothing algorithm for the estimation of parameters in neuronal models is presented. The algorithm constitutes a useful addition to the computational neuroscience toolbox and an alternative to typically data-hungry spectral and information-theoretic methods. In particular, the algorithm has been designed to perform well for noisy data sets consists of a relatively short time-series (perhaps only a few periods observed), data is completely missing on one or more variables (latency), and model mis-specification is almost certain (simplifying assumptions). Simulation studies using Wedge estimation are performed for some well-known neuronal oscillators. The posterior coverage intervals from these studies support the recommendation that the Wedge algorithm is appropriate for hypothesis generation and exploration but should be used with caution when the goals of analysis are inferential. |
|---|
| Coauthors | Johannes Reisert, Haiqing Zhao |
|---|
| Location | Woodward Lobby (Monday-Tuesday) |
|---|
