Some of the most productive scientific collaborations arise when people from wildly different backgrounds meet by chance and bring new, fresh approaches that resist traditional techniques. Making people from different backgrounds mix, however, is a challenging task, which is typically just left to serendipity.
We describe an approach jointly developed by a Cambridge and Trento team to favour the mixing of scientists at conferences. Our approach is inspired by 'speed dating' - where pairs of scientists are formed and briefly discuss their work for 5 minutes, and then exchange information.
However, instead of leaving the formation of pairs to chance, we develop a framework based on network theory to create a compatibility matrix between all possible pairs of scientist (based on different criteria, such as maximizing heterogeneity, or minimizing overlap between known fields, etc) and then use the Maximum-Weight Perfect Matching Algorithm to find the optimal pairs.
We show how to solve this using tools from the SciPy ecosystem (pandas, networkx, numpy, matplotlib) and describe an open source implementation of our algorithm.
