Small experiment in computing the ground state of the 1d Ising/Heisenberg/J1J2 model using neural networks. Both convnets and lstms are implemented. Built using Jax and Flax.
For usage instructions try out the notebook.
Follows broadly in the steps of:
Recurrent Neural Network Wavefunctions
Iterative Retraining of Quantum Spin Models using Recurrent Neural Networks
Deep autoregressive models for the efficient variational simulation of many-body quantum systems