Here is an overview figure describing the method. Huge thank you to my co-authors @frmetz.bsky.social and @gppcarleo.bsky.social for all the help and feedback on this work! 🧵6/6

On the other hand, the quantum circuit increases the expressivity of the classical ansatz - allowing for the simulation of dynamics with few parameters. This helps making the evolution using TDVP more stable.🧵5/6

Importantly: The circuit contains no variational parameters, any derivatives can be computed purely classically using backpropagation. We show that a simple Jastrow ansatz can improve the fidelity of Trotterized quantum dynamics of spin systems. 🧵4/6

💻 Given the samples from the quantum device, we can construct the equations of motion to update the parameters in the classical ansatz in order to correct for the approximation of the Hamiltonian. For this, we extended TDVP for explicitly time-dependent ansätze.🧵3/6

⚛️On the quantum device, we evolve an initial state in time according to an approximation of the target Hamiltonian. This scenario arises when coarse Trotterization or a hardware efficient simplification of the Hamiltonian is used. This circuit is then sampled at each time step in multiple bases.🧵2/6