Neural Network Connectome Mapping

This is a collaboration between Prof. Martin Vetterli and Dr. Amin Karbasi (Yale).

This project is about inferring the connectivity of neural graphs, at different levels (e.g., synaptic, micro-circuit, functional, etc.), by merely observing the (firing) activity of nodes (neurons/circuits/regions). More specifically, we are given the activity of a set of nodes in a graph and are asked to do a “reverse engineering” and infer the underlying directed connectivity between the nodes. While the problem is very similar to what we have in some other areas, e.g., social networks, the complexity in behaviour of neurons as well as the dynamics of the neuronal graphs make the problem more challenging. In particular, the integrating nature of the neurons plus the essential delay on inter-neuronal connections make the inference task particularly difficult. In this project, we develop novel techniques and provide theoretical bounds on the performance of the inference algorithm in certain reasonably realistic scenarios. We also plan to improve the accuracy of the algorithms for practical situations and apply the developed methods to real recorded data. The results would hopefully provide some insight about the functionality of different regions in animals/humans brain as well as how ageing and learning affect the connectivity in certain parts.

Related Papers
Learning neural connectivity from firing activity: efficient algorithms with provable guarantees on topology
A. Karbasi, A. H. Salavati, M. Vetterli
Learning network structures from firing patterns
A. Karbasi, A. H. Salavati, M. Vetteri

Some Useful Resources