DeepMind has open-sourced the Fermionic Neural Network (FermiNet) designed to facilitate research in computational physics and chemistry by enabling users to model the quantum state of large sets of electrons. FermiNet was developed to approximate solutions to the many-electron Schrodinger equation without using external data. To approximate quantum wavefunctions efficiently and accurately, DeepMind researchers designed FermiNet to be antisymmetric with respect to its inputs. Because FermINet can generate samples from the wavefunction directly, it does not require additional inputs for training.