Additional group theory functions for error correction #351

IsaacP1234 · 2024-08-30T19:51:36Z

Adding logical_operator_canonicalize, commutavise, and embed functions, as well as SubsystemCodeTableau datastructure and minor performance improvements for groupify and normalizer.

Fe-r-oz · 2024-09-04T20:28:09Z

Hi, Isaac,

The merge conflicts are occuring because the commits from the top to 39af6f0 are actually from PR #293 and these commits are being pushed again. If you submit the PR of commits starting from (add LOC function)109104e to the end, there would be no merge conflict.

You should request Kenneth to fork the latest master branch of QC.jl, then make a new branch, say newgrouptab. Then, you can fork newgrouptab, and do commits and push PR. It seems here you have forked the grouptableaux branch of Kenneth whose first commit is part of PR 293 (grouptableaux). Then you made loc-comm-embed by making a new branch based on grouptableaux and the pushing your commits to loc-comm-embed . But the first branch (grouptableaux) does not coincide with the master that has grouptableaux.jl and it's needs to be updated, and hence all the commits have reappeared here.

IsaacP1234 · 2024-09-04T21:30:42Z

Thank you for your explanation and solution Feroz.

@KDGoodenough when you get chance, could you fork the latest version of QuantumClifford and create a branch that I can fork as described by Feroz? I can also fork the repository myself if that would make more sense @Krastanov

Krastanov · 2024-09-05T00:46:03Z

This is a bit roundabout and unnecessary.

Isaac, if you do not want to bother with git, just make a new branch from the latest master branch and copy over your code, make sure it is your latest implementation and submit a new PR.

If you prefer to do it the "proper way", make a new branch from the latest master branch and cherry pick ("cherrypick" is the name of a git command) the commits from this branch that you care about and submit a new PR.

Another "proper way" would be to go back to a clean commit of yours and use an interactive rebase. Your git history is a bit messy so this would probably be a bit difficult.

There is no need for Kenneth to make new forks of anything. If you do want to follow Feroz's suggested approach, all the branches and commits referred to in his post are public so you can do all the steps yourself. But that is just one out of many ways you can do this.

Lastly, you can also simply fix the merge conflicts here without doing cherrypicks or rebases or anything like that. The merge conflicts seem reasonably small, so this solution might be easiest.

IsaacP1234 · 2024-09-06T04:28:48Z

Thank you, I resolved the conflicts here on Github.

I will write a changelog when review is complete.

Additionally, for the SubsystemCodeTableau datastructure, what should (de)stabilizerview and logicalx/zveiw return when their respective category is empty? For example, when the result of logical_operator_canaonicalize contains no operators that commute with all other operators, what should (de)stabilizerview return? Currently, I just have them return an n-qubit PauliOperator.

KDGoodenough · 2024-09-06T04:39:59Z

For logical_operator_canonicalize, note there always exist one operator that commutes with everything, no matter the input --- the identity operator. For the other cases ((de)stabilizerview and logicalx/zview), one can make a similar argument, I believe.

Krastanov · 2024-09-06T12:25:43Z

For the stabilizer part of the tableau (the part that commutes with the logical views), are you returning a generating set or all elements of the group? If it is a generating set, I would suggest keeping the same conventions as in MixedDestabilizer and just return a tableaux with zero rows:

julia> s = MixedDestabilizer(S"XX ZZ") # A full rank tableau
𝒟ℯ𝓈𝓉𝒶𝒷
+ Z_
+ _X
𝒮𝓉𝒶𝒷
+ XX
+ ZZ

julia> logicalxview(s)


julia> logicalxview(s) |> typeof
Stabilizer{QuantumClifford.Tableau{SubArray{UInt8, 1, Vector{UInt8}, Tuple{UnitRange{Int64}}, true}, SubArray{UInt64, 2, Matrix{UInt64}, Tuple{Base.Slice{Base.OneTo{Int64}}, UnitRange{Int64}}, true}}}

julia> length(logicalxview(s))
0

IsaacP1234 · 2024-09-08T00:39:39Z

I think the tests are failing for reasons unrelated to my code. The Julia-1/Ubuntu tests fail when testing noisy circuits, which don't seem to overlap with my code, and the jet tests are classified as 'broken', but it doesn't seem like it involves my code based on the output. the Buildkite fails on the downstream tests on a test labeled "piracy", which also seems separate from my code.

If any of these errors are related, let me know and I will fix them. If not, the code is ready for review.

Krastanov

I did a quick first round of review. I have a couple of minor comments on the code, but most of my comments are on documentation. The implementation looks great, and it seems it will be pretty useful for Kenneth's work. Some work still to do on making sure that other people can use this by reading the documentation (maybe Kenneth can help with some of the docs).

Krastanov · 2024-09-08T18:58:36Z

src/QuantumClifford.jl

@@ -76,6 +76,7 @@ export
 graphstate, graphstate!, graph_gatesequence, graph_gate,
 # Group theory tools
 groupify, minimal_generating_set, pauligroup, normalizer, centralizer, contractor, delete_columns,
+ logical_operator_canonicalize, commutavise, embed, SubsystemCodeTableau,


@KDGoodenough , @IsaacP1234

Any preferences between commutavise and commutivise? Is this term used elsewhere by anyone else? Where does the name come frome?

Any other suggestions for logical_operator_canonicalize? The name is a bit long and I am not quite sure whether it conveys what is happening well.

embed as a name is already used for other purposes so this also would need to be renamed. There is already embed methods which embed a small unitary operator in a larger space by padding it with identities.

let's use the following name for logical_operator_canoninicalize to keep in similar to the other canonicalize methods we have:

canonicalize_noncomm

let's, for now, change commutavise to commutify to keep it in sync with simplify or lambdify or sympify

let's, for now, rename embed to matroid_parent

Krastanov · 2024-09-08T19:00:16Z

src/grouptableaux.jl

+A tableau representation of the logical operator canonical form of a set of Paulis.
+The index of the first operator that commutes with all others is tracked so that the
+stabilizer, destabilizer, logical X, and logical Z aspects of the tableau can be 
+represented.


I do not think logical operator canonical form is a known thing in the field. Could we document this a bit more, with definitions and explanations for why this is interesting and useful, maybe a few doctest examples?

Check with Kenneth on this, maybe ask him to help with the documentation for it.

Krastanov · 2024-09-08T19:00:58Z

src/grouptableaux.jl

+represented.
+"""
+mutable struct SubsystemCodeTableau <: AbstractStabilizer
+ tab::Tableau


Parameterize the tableau type (same as MixedDestabilizer for instance), to ensure type information is not lost here.

Krastanov · 2024-09-08T19:01:30Z

src/grouptableaux.jl

+ return SubsystemCodeTableau(tab, index, length(s), m, (index-1)/2)
+
+end


Suggested change

return SubsystemCodeTableau(tab, index, length(s), m, (index-1)/2)

end

return SubsystemCodeTableau(tab, index, length(s), m, (index-1)/2)

end

Krastanov · 2024-09-08T19:03:41Z

src/grouptableaux.jl

+
+end
+
+Base.length(t::SubsystemCodeTableau) = length(t.tab)


Is this used somewhere? Given the complicated internal structure, I am reluctant to define a "homogenous" length function as I am not sure what it would mean semantically. If it is not used, let's delete it.

Krastanov · 2024-09-08T19:42:57Z

src/grouptableaux.jl

+ com, d1= QuantumClifford.commutavise(t)
+ norm = QuantumClifford.normalizer(com.tab)
+ state, d2 = QuantumClifford.commutavise(norm)


Suggested change

com, d1= QuantumClifford.commutavise(t)

norm = QuantumClifford.normalizer(com.tab)

state, d2 = QuantumClifford.commutavise(norm)

com, d1= commutavise(t)

norm = normalizer(com.tab)

state, d2 = commutavise(norm)

Krastanov · 2024-09-08T19:43:29Z

src/grouptableaux.jl

 """
 Return the full Pauli group of a given length. Phases are ignored by default, 
-but can be included by setting `phases=true`.
+but can be included by setting `phases = true`.


Suggested change

but can be included by setting `phases = true`.

but can be included by setting `phases=true`.

Krastanov · 2024-09-08T19:43:51Z

src/grouptableaux.jl

 end
 end
-
- return Tableau(normalizer)
+ return QuantumClifford.Tableau(norm)


Suggested change

return QuantumClifford.Tableau(norm)

return Tableau(norm)

Krastanov · 2024-09-08T19:44:17Z

test/test_group_tableaux.jl

@@ -16,34 +16,75 @@
 s_test = copy(s)


@KDGoodenough , could you check that this file makes sense (the test file)?

T_str is already public, and Tableau is not needed to be public for this code. I can still make it public if you want.

oh, no need to make Tableau public then, but in various examples switch to @T_str instead of the more complicated construction of Tableau that you are using.

src/grouptableaux.jl

Co-authored-by: Stefan Krastanov <github.acc@krastanov.org>

KDGoodenough · 2024-09-10T12:44:25Z

@IsaacP1234, could you add a reference of this paper to what we now call logical_operator_canonicalize?

https://scholar.google.com/scholar?q=https://arxiv.org/pdf/1302.3428&hl=en

KDGoodenough · 2024-09-10T13:12:51Z

And a reference to this paper and Fig. 1: https://arxiv.org/pdf/2406.02427 for the delete, contractor, and embed function please!

IsaacP1234 · 2024-09-12T02:40:33Z

@KDGoodenough I have rewritten the documentation for canonicalize noncomm, commutify, and matroid_parent to be more clear, but is there anything about the functions or the theory behind them that I should include in the documentation?

…mClifford.jl into loc-comm-embed fixed merge conflict with origin

IsaacP1234 · 2024-09-15T20:53:16Z

I think I have completed all the requested changes, but I am not aware of more uses of commutify aside from matroid parent to add to its documentation. @KDGoodenough are there any I should add?

Additionally, should I add the newly-referenced papers to references.md?

KDGoodenough · 2024-09-16T05:25:27Z

@IsaacP1234 I think it's good like this! There's not too much work (yet!) that uses this commutify business. My guess would be that yes, it would be good to add these references to the .md file as well, but Stefan would know better.

IsaacP1234 · 2024-09-16T23:57:45Z

Changelog:

New error correction tools:
- canonicalize_noncomm function to find a generating set with minimal anticommutivity
- SubsystemCodeTableau data structure to represent the output of canonicalize_noncomm
- commutify function to find a commutative version of a non-commutative set of Paulis with
  minimal changes
- matroid_parent function to, for set of Paulis that doesn't represent a state, find a version
  that does.

KDGoodenough and others added 30 commits March 27, 2024 21:12

first commit

3ec3098

added GT dependecies to project.toml

dee1c59

started GT test file

a708c00

started groupify tests

8627fee

tests for groupify/generating set, pauli ops don't work

3bf54ae

removing unnecessary dependencies

4f377b7

removed print statements

929607d

completed get_gen_set tests

1161ff2

added phase arg to pualigroup and tests for normalize

da32a75

added tests for center, fixed center

09ea8b5

formatted tests

e2af6b7

added changes to changelog

c1510a0

removed SafeTestSets dependency, formatted changelog

f22cf6e

fixed center function and pauligroup phases

5ded49b

added tests for logical_op_canonicalize

59cf3be

updated LOC tests

2076a52

updated LOC tests and CHANGELOG

b7d0eab

updated GGS+tests to work with I operators

7694800

added method descriptions

bd5e63a

removed print statements + fixed center tests

d4265dd

fixed changelog grammar

b9071e1

formatted+commented unfinished functions

43f4736

fixed changelog

afa2d17

completed changelog

c1282d8

removed unnecessary comments

79f8f75

moved pauli delete to p_o.jl

0fc1725

removed changelog update to prevent merge conflicts

6029b7f

converted delete to deleteat! and moved to paulioperator.jl

d2bf12a

finished contract, cleaned up code, added documentation

b633615

added tests for contract

be86915

fix qc.jl merge conflict

a1c84b7

IsaacP1234 and others added 3 commits September 6, 2024 00:12

fully implemented SSCT

688ca8c

Merge branch 'master' into loc-comm-embed

5cfc2c9

spell check

a4b73d2

IsaacP1234 added 3 commits September 6, 2024 18:10

fixed view functions+allowed identity operators

b199101

fixed doctest

5b5b45b

Fixing doctests

3d2f1d5

Krastanov requested changes Sep 8, 2024

View reviewed changes

remove unnecessary qcdot

903d843

Co-authored-by: Stefan Krastanov <github.acc@krastanov.org>

IsaacP1234 added 2 commits September 11, 2024 19:56

renamed+changed docs+cleaned up

50fea07

fixed doctests+added citations.

5500322

IsaacP1234 added 3 commits September 11, 2024 22:43

Merge branch 'loc-comm-embed' of https://github.com/IsaacP1234/Quantu…

04c4c29

…mClifford.jl into loc-comm-embed fixed merge conflict with origin

improved docs+added referencese

5257f1a

updated SSCT docs

64e1caa

IsaacP1234 added 2 commits September 16, 2024 14:14

fixed citing

ee7beb3

test error fix

c785372

IsaacP1234 requested a review from Krastanov September 17, 2024 00:01

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Additional group theory functions for error correction #351

Additional group theory functions for error correction #351

IsaacP1234 commented Aug 30, 2024

Fe-r-oz commented Sep 4, 2024

IsaacP1234 commented Sep 4, 2024

Krastanov commented Sep 5, 2024

IsaacP1234 commented Sep 6, 2024

KDGoodenough commented Sep 6, 2024

Krastanov commented Sep 6, 2024

IsaacP1234 commented Sep 8, 2024

Krastanov left a comment

Krastanov Sep 8, 2024

Krastanov Sep 10, 2024

Krastanov Sep 10, 2024

Krastanov Sep 10, 2024

Krastanov Sep 8, 2024

Krastanov Sep 8, 2024

Krastanov Sep 8, 2024

Krastanov Sep 8, 2024

Krastanov Sep 8, 2024

Krastanov Sep 8, 2024

Krastanov Sep 8, 2024

Krastanov Sep 8, 2024

IsaacP1234 Sep 12, 2024

Krastanov Sep 12, 2024

KDGoodenough commented Sep 10, 2024

KDGoodenough commented Sep 10, 2024 •

edited

Loading

IsaacP1234 commented Sep 12, 2024

IsaacP1234 commented Sep 15, 2024

KDGoodenough commented Sep 16, 2024

IsaacP1234 commented Sep 16, 2024

		return SubsystemCodeTableau(tab, index, length(s), m, (index-1)/2)

		end

	but can be included by setting `phases = true`.
	but can be included by setting `phases=true`.

Additional group theory functions for error correction #351

Are you sure you want to change the base?

Additional group theory functions for error correction #351

Conversation

IsaacP1234 commented Aug 30, 2024

Fe-r-oz commented Sep 4, 2024

IsaacP1234 commented Sep 4, 2024

Krastanov commented Sep 5, 2024

IsaacP1234 commented Sep 6, 2024

KDGoodenough commented Sep 6, 2024

Krastanov commented Sep 6, 2024

IsaacP1234 commented Sep 8, 2024

Krastanov left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

KDGoodenough commented Sep 10, 2024

KDGoodenough commented Sep 10, 2024 • edited Loading

IsaacP1234 commented Sep 12, 2024

IsaacP1234 commented Sep 15, 2024

KDGoodenough commented Sep 16, 2024

IsaacP1234 commented Sep 16, 2024

KDGoodenough commented Sep 10, 2024 •

edited

Loading