ppravatto / binary-vqe Goto Github PK

Add a post-processing option to search correlations between the VQE optimization starting point and the endpoint in a multiple VQE scan.

Can the optimized parameters of a N qubits calculation be used as a guess in a N+1 qubits calculation? If yes it can help to mitigate the error on higher qubit count problems.

The iterations should be saved from a proper callback function (if available) and not from the optimization target function

Per facilitare il confronto della convergenza usando ottimizzatori diversi o entangler diversi o shots diversi si può pensare di rinominare i file di output con un codice che indichi univocamente i parametri usati.
Es:
CL20k potrebbe indicare un calcolo svolto con il Cobyla linear entangler 20 mila shots

The program uses a relative path to refer to files. If an absolute path is used instead the program can be called from other folders and operate on files in different locations.

The current implementation requires the user to choose a set of basis functions that fully utilize the qubit register. If a different number is selected an undefined behaviour is encountered. If the qubit register is not fully utilized non-physical states can appear in the wavefunction. These states do not appear in the Hamiltonian matrix and correspond to null eigenvalues that drives the variational procedure to zero (if the variational form allows it). These states need to be eliminated by offsetting them or the number of allowed basis functions should be hardcoded.

RyRz entangler can be condensed into few gates removing barriers and converting Ry, Rz, H operations to U3 gates. Transpile can possibly be used but topology driven swaps can complicate the software structure.