Date of Submission

Spring 2018

Academic Programs and Concentrations

Biology

Project Advisor 1

Arseny Khakhalin

Project Advisor 2

Kerri-Ann Norton

Abstract/Artist's Statement

One of the major areas of research in computational neuroscience is focused on inferring the connections within populations of neurons from the signaling activity of these populations. Methods of reconstructing structural neuronal network connectivity are limited and, in large populations, technically infeasible. Current methods that reconstruct networks of large populations relate connectivity to calcium imaging recordings of these networks. Here, we introduce a machine-learning approach to inferring connectivity from spike-time data extracted from calcium imaging recordings. First, we simulate populations of neurons with the NEST simulator to produce downsampled spike trains. We develop a model based on neural networks, which is a widely applied machine-learning method. The model is updated with gradient descent on the error via backpropagation, and the performance is compared to the widely used cross-correlation method of extracting functional connectivity. We then train the models on simulated data and in-vivo calcium imaging data from Xenopus Laevis tadpoles.

Open Access Agreement

On-Campus only

Creative Commons License

Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.

Share

COinS