Post Selection on StateVector

#1

Dear All,

First of all let me congratulate with the entire team, I haven’t used Perceval in a long while and I have seen a tremendous progress!

I have a comment/question regarding the post selection functionality.
Currently is it possible to apply a PostSelect function only to “erase” unwanted states from a ProbabilityDistribution?
In that case, do you plan to extend it to StateVector? A general postselection rule allows to throw away unwanted portions of any kind of states, potentially, maintaining its coherence.

As a result, the Simulator object despite evolving a state cannot really apply a rule to the output, despite the presence of the set_postselection method (I mean, I can still get transform to distribution and postselect there but I’d like to keep the state coherent, along with its phasis!).

Thanks a lot & best

#2

Hello,

Thanks for your kind words, I’ll pass them to the whole software team! :slight_smile:

Currently a PostSelect object can only do one thing: you feed him a Fock state and if answers True (select it) or False (filter it out). It’s used inside the Processor class to post-select measured states during a probability distribution or a sampling simulation.

So, in Perceval 0.10, you’d have to add some code in order to apply post-selection to an evolved StateVector. For instance, call:

def postselect_statevector(ps: PostSelect, sv: StateVector):
    sv_out = StateVector()
    for state, ampli in sv:
        if ps(state):
           sv_out += ampli*state
    sv_out.normalize()
    return sv_out

This will be done automatically in Perceval 0.11.

I hope this helps!
Best wishes,
Eric

1 Like
#3

Hi Eric,

thanks for your answer and the code example, it was really helpful for me! I can only confirm what FGiorgino already said, it is really nice to work with perceval!

One thing that I would like to point your attention to: The added functionality of PostSelect in perceval 0.11, that is doing what you posted before, is (at least for me) not working. I think there is a tiny error in the code.

When I run the following code (I am using python 3.9.20 on a windows machine, [MSC v.1929 64 bit (AMD64)], perceval version 0.11.0):

import perceval as pcvl

state_vec = pcvl.StateVector("|0,1>")+pcvl.StateVector("|1,0>")

post_sel = pcvl.PostSelect("[0] == 1")
post_sel(state_vec)

the following error is raised:

post_sel(state_vec)

  File ~\Anaconda3\lib\site-packages\perceval\utils\postselect.py:127 in __call__
    s += state[i]

TypeError: unsupported operand type(s) for +=: 'int' and 'exqalibur.FockState'

I also checked it in jupyter and there a bit more of the definition of the called function is shown:

125 s = 0
    126 for i in indexes:
--> 127     s += state[i]
    128 if not operator(s, value):
    129     return False

I think the problem is that different to your code example, where you did

sv_out = StateVector()

here the function does

 s = 0

and therefore the problem with the addition between an int and a StateVector object arises.

For me it was not a problem, I just defined a function myself following your code example, but it might be nice for others to correct this error.

Best,
Lukas

#4

Hi Lukas,

Starting with Perceval 0.11, you can use the post_select_statevector function which is available in perceval.utils. With it, you can use any combination of post-selection conditions and heralds to drop all basic states that do not fit in you state vectors.

Cheers,
Eric

1 Like