PredictLand participates in a research with Machine Learning techniques in collaboration with a research institute of the Consejo Superior de Investigaciones Científicas.
Customer
INMA(Instituto de Nanociencia y Materiales de Aragón) is a joint research institute between CSIC (Consejo Superior de Investigaciones Científicas) and Unizar (Universidad de Zaragoza). It is a world reference in areas ranging from materials science to quantum computing.
Challenge
The challenge we set ourselves in the collaboration was to find the discrete symmetries of a physical system with machine learning based techniques. Finding such symmetries is of paramount importance for two reasons:
- The discrete symmetries of a physical system give much information about the system; for example, they influence how the system interacts with light.
- To date, finding the symmetries of an unknown physical system was a highly non-trivial task.
Solution
As a result, we designed a neural network with a novel architecture such that, from observational data of the physical system in question, it is able to automatically find how many symmetries the system has and what these symmetries are: whether rotational, translational, etc.
The results are posted at https://arxiv.org/abs/2307.13457.
