Fig. 1: Pictorial representation and applications of Foundation Neural-Network Quantum States.
From: Foundation neural-networks quantum states as a unified Ansatz for multiple hamiltonians
The panel (a) shows a pictorial representation of Foundation Neural-Network Quantum States (FNQS), which, unlike traditional NQS, process multimodal inputs by incorporating both physical configurations and Hamiltonian couplings to define a variational wave function amplitude over their joint space. FNQS enable a range of applications, including the efficient simulation of disordered systems [see panel (b)] and the estimation of the quantum geometric tensor in coupling space, also known as the fidelity susceptibility, for the unsupervised detection of quantum phase transitions [see panel (c)]. Moreover, FNQS combined with the public availability of the architectures allows users to leverage pretrained models to explore coupling regimes beyond those encountered during training [see panel (d)].
