Closed zmorrell closed 2 years ago
Merging #35 (2a90105) into main (d7adf1b) will increase coverage by
0.05%
. The diff coverage is100.00%
.
@@ Coverage Diff @@
## main #35 +/- ##
==========================================
+ Coverage 92.80% 92.86% +0.05%
==========================================
Files 5 5
Lines 598 603 +5
==========================================
+ Hits 555 560 +5
Misses 43 43
Impacted Files | Coverage Δ | |
---|---|---|
src/base.jl | 98.33% <100.00%> (+0.15%) |
:arrow_up: |
:mega: Codecov can now indicate which changes are the most critical in Pull Requests. Learn more
I didn't get a deep understanding of what Marc had in mind for this feature but I had imagine the API would work for an arbitrary Hamiltonian. I was expecting it would look something like z_probailities_boltzmann(hamiltonian_transverse_ising(...),1.0)
. My inclination is to make the beta parameter explicit as this is a fundamental term in the Boltzmann model. We might even break it down further like, z_measure_probabilities(distribution_boltzmann(hamiltonian_transverse_ising(...),1.0))
In general I would also reserve the term "sample" for functions that return a vector/histogram of states, rather than ones that can return exact probabilities.
@ccoffrin I agree that the word sample
has some different meanings when it comes to statistics, so it would probably be a good idea for me to change the name. If I am not mistaken, the code as it is written will work for any Hamiltonian that is a function of s
. I will send you an email separately pointing you to what I believe Marc had in mind. I will change the name to the suggested z_probabilities_boltzmann
suggestion since I think that name is consistent with the rest of the package.
@ccoffrin The testing on this is rather rudimentary, but I couldn't figure out a much better way to do it, since theoretically boltzmann sampling of the classical and quantum hamiltonian would give the same results under the tests which I added. I think it is best to just double check that the code is correct, which should be fine since it is only 4 lines or so.