Inspired by our work in network inference, we teamed with a group of Machine Learning specialists and launched the first Chalearn Connectomics Challenge in the Kaggle platform. The challenge was aimed at data scientists and we provided a set of simulated calcium imaging recordings from neuronal cultures, similar to the ones we used for our previous work, and the goal was to infer the network topological structure from the recordings (the participants only had access to fluorescence and positional data and some training datasets).
The challenge attracted almost 150 teams and generated an intense, yet cooperative competition and over 20 teams were able to surpass our benchmark score based on Transfer Entropy. All the winning teams made their code open source and available, and we are currently in the process of analyzing the results and their robustness. The different approaches taken by the participants and their codes will hopefuly be of use in future connectomics research.
You can find more information in this proceeding (while we prepare the final paper).