“Open sourced neural network system to realize quantum chemistry calculation”

DeepMind, a developer of Go AI AlphaGo, etc., announced FermiNet, a system that uses quantum chemical computation as a neural network, in September 2020, while also announcing the open source Ferminet.

Ferminet explains that it is a high-precision neural network that demonstrates a deep learning system that calculates the principle of atomic and molecular energy. It is a system that is expected to be applied in various fields as it has the potential to help not only solve basic scientific problems, but also to computerize new materials and chemical synthesis prototypes in the future.

The research team decided to open-source Ferminet so that it could be applied to various scientific fields. The Ferminet source code is available on GitHub . In response, the research team said that Ferminet is the beginning of a great achievement towards fusion of deep learning and quantum chemical computation, and hopes that it will stimulate the idea of a better new network architecture. Several groups are already using Fermine for actual research.

In addition to quantum chemistry calculations, the research team revealed that it is expected to be applied to difficult problems in materials science and material physics. By releasing the source code used in the experiment, he said he expects other researchers to be applied to new applications based on their research. Related information can be found here .