Yael Ben-Haim

Yes I know, but I think there is another piece of code for the last spreadsheet.

Based on the notebooks, I created the following code: ``` from qiskit.providers.aer import QasmSimulator from qiskit.optimization import QuadraticProgram from qiskit.optimization.algorithms import MinimumEigenOptimizer from qiskit.aqua.algorithms import QAOA from qiskit.aqua.components.optimizers import SLSQP...

I'm able to restore the results for 20 qubits, add see the difference between the simulators. I'm checking it now.

Here's an update. - Some of the differences are because in the QASM notebook, the options in the `set_options` line (of the optimizer) are different from the MPS notebook. -...

I understand now what's going on. The bottom line is a numerical difference 10 positions after the decimal point, which propagates to totally change the flow of QAOA. Note that...

I think it's best to consult with someone from Aqua about the best way to use QAOA (for example, what is the recommended number of shots?). Also, following discussion in...

Although `include_custom` is `True`, when I debug I see a circuit without snapshots, executed 1024 times.

The `O(s + 2^n)` algorithm should be implemented at some point, because a sub-optimal code is just not nice. However the gain in performance compared to the `O(sn + 2^n)`...

If generating samples from the histogram takes `O(s^2)` (I feel that it can be improved to `O(s log(s))` and maybe more) then possibly #819 just has smaller complexity, at least...

Samples from histogram is just `O(s)`, it's a random permutation.