Don't add Hadamards to a user-defined initial state in QAOA #1362
Conversation
Milos9304
requested review from
levbishop,
manoelmarques and
woodsp-ibm
as code owners
|
|
Milos9304
changed the title
Milos9304
changed the title
|
Hi @Milos9304, thanks for the contribution! Could you also add a release note saying that you fixed this? 🙂 (And a test of course, but you already mentioned you're working on that) |
Cryoris
changed the title
Cryoris
added
the
Changelog: Bugfix
… release note for the bugfix.
Done. I am quite confused with some of the unrelated tests failing for weird errors like e.g. ECONNRESET etc. My test in P.S. This is my very first open-source contribution so I apologize if I am doing some mess around here. |
|
The build system caches the downloaded packages in order to save time on subsequent builds. It seems there was a problem saving the cache. I'll update the branch since the master has had changes done recently and this will start the build again. Sometimes there are problems with the build that are unrelated to what you have done - I can understand how that can be confusing when you first see such things failing! |
manoelmarques
added
the
stable backport potential
…ommunity#1362) * Fixing QAOA initial_state problem * fixing code style warnings * Creating QAOA-initial_state-dont-mix-with-Hadamards bugfix test and a release note for the bugfix. * Removing code that appeared in pull request due to error * Fixing import math error * Fixing code style issues. Co-authored-by: Steve Wood <40241007+woodsp-ibm@users.noreply.github.com>
…ommunity/qiskit-aqua#1362) * Fixing QAOA initial_state problem * fixing code style warnings * Creating QAOA-initial_state-dont-mix-with-Hadamards bugfix test and a release note for the bugfix. * Removing code that appeared in pull request due to error * Fixing import math error * Fixing code style issues. Co-authored-by: Steve Wood <40241007+woodsp-ibm@users.noreply.github.com>
…ommunity/qiskit-aqua#1362) * Fixing QAOA initial_state problem * fixing code style warnings * Creating QAOA-initial_state-dont-mix-with-Hadamards bugfix test and a release note for the bugfix. * Removing code that appeared in pull request due to error * Fixing import math error * Fixing code style issues. Co-authored-by: Steve Wood <40241007+woodsp-ibm@users.noreply.github.com>
…ommunity/qiskit-aqua#1362) * Fixing QAOA initial_state problem * fixing code style warnings * Creating QAOA-initial_state-dont-mix-with-Hadamards bugfix test and a release note for the bugfix. * Removing code that appeared in pull request due to error * Fixing import math error * Fixing code style issues. Co-authored-by: Steve Wood <40241007+woodsp-ibm@users.noreply.github.com>
Summary
initial_stateprovided to QAOA gets composed with Hadamard gates, which is an undesired behaviour. I assume this bug is a result of implementation of the case where noinitial_stateis set. In this default case the input state to ansatz circuit is |+++...++> as described in QAOA Farhi's paper, hence the Hadamards.Details and comments
My solution is straightforward and can be easily seen from code. I am working on fixing the tests.