Exploiting the Tree Structure of Neurons: Directed Tree Neural Networks
The brain is structured into various areas that each contain different types of neurons. Clustering of such cortical neurons by their shape (including its branching structure) has a long tradition in biology research. It is known that the shape of a neuron is tightly connected to its function in the brain. In the last couple of years, the amount of data has risen massively such that manual inspection is no longer feasible. We are thus interested in purely data-driven methods to cluster such neurons. Unsupervised or semi-supervised techniques could allow to make use of all the available data and provide high-accuracy clusterings and classifications without relying (too much) on domain knowledge.
The aim of this project is to exploit the unique structure of neurons - namely, the dendrites form a forest (or a tree if we use the some as a common root node). So far, all techniques for embedding the shape of neurons tried to infer information about the structure of the neuron based on techniques that work for all kinds of graphs, in this project we want to change that.
Note that the literature around directed GNNs focuses mostly on node embedding and in general is rather sparse (compared to general GNN literature) and the project is thus about adapting GNNs and directed GNNs to trees or developing entirely new models that directly exploit that a neuron can be represented as a rooted tree.
Requirements
- Python programming
- Interest in Graphs and other kinds of structured data
- Interest in working with neurons and biological data
- Experience with PyTorch and/or Tensorflow is a plus
Literature
- PyTorch Geometric Signed Directed: A Software Package on Graph Neural Networks for Signed and Directed Graphs arxiv
- SDGNN: Learning Node Representation for Signed Directed Networks (AAAI 2021)
- edGNN: a Simple and Powerful GNN for Directed Labeled Graphs arxiv
