New feature: Global phase linear constraints #62

szabino-mrszabo · 2022-10-04T23:16:06Z

No description provided.

…and set target as the phaseless target

codecov · 2022-10-04T23:59:41Z

Codecov Report

Base: 96.05% // Head: 96.13% // Increases project coverage by +0.07% 🎉

Coverage data is based on head (53ff494) compared to base (381b589).
Patch coverage: 94.87% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #62      +/-   ##
==========================================
+ Coverage   96.05%   96.13%   +0.07%     
==========================================
  Files          13       13              
  Lines        1876     1939      +63     
==========================================
+ Hits         1802     1864      +62     
- Misses         74       75       +1

Impacted Files	Coverage Δ
src/data.jl	`92.37% <84.61%> (-0.11%)`	⬇️
src/utility.jl	`95.12% <93.54%> (-0.13%)`	⬇️
src/constraints.jl	`98.05% <100.00%> (+0.14%)`	⬆️
src/log.jl	`95.65% <100.00%> (+1.57%)`	⬆️
src/objective.jl	`96.87% <100.00%> (ø)`

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

harshangrjn

Added comments

harshangrjn · 2022-10-05T16:16:57Z

src/utility.jl

+    ref_nonzero_r = 0
+    ref_nonzero_c = 0
+    if data["are_gates_real"]
+        for i=1:n_r, j=1:n_c


Lines 728-734 applies only for all-real gates, which is not covered in the test. Could you add a test to cover this within "QC_model Tests: Global phase constraints"

added unit test to cover all-real case as well

harshangrjn · 2022-10-05T16:18:38Z

src/utility.jl

+    return ref_nonzero_r, ref_nonzero_c 
+end
+
+function _get_ref_nonzero_index_of_original_target(T::Array{Complex{Float64},2})


This function isn't covered in the tests. Do we need this separately from the previous function _get_ref_nonzero_index_of_real_target. If possible, I would combine them.

I added a unit test for this as well.

I see where you are coming from; I could add an option to make it one function but I feel it is probably easier to understand to have 2 separate functions for 2 things that I consider very different objects. Like apples and pears.

One is just a complex array 'target' as it is an input

One is a QuantumCircuitModel.data.target_gate thing which is under the hood a real array but nonetheless is this part of this magical type QuantumCircuitModel

let me know what you think.

harshangrjn · 2022-10-05T16:22:02Z

docs/src/quickguide.md

@@ -46,7 +46,7 @@ QCOpt takes two types of user-defined input specifications. The first type of in
 | `elementary_gates` | Vector of all one and two qubit elementary gates. The menagerie of quantum gates currently supported in QCOpt can be found in [gates.jl](https://github.com/harshangrjn/QuantumCircuitOpt.jl/blob/master/src/gates.jl). |
 | `target_gate` | Target unitary gate which you wish to decompose using the above-mentioned `elementary_gates`.|
 | `objective` | Choose one of the following: (a) `minimize_depth`, which minimizes the total number of one- and two-qubit gates. For this option, include `Identity` matrix in the above-mentioned `elementary_gates`, (b) `minimize_cnot`, which minimizes the number of CNOT gates in the decomposition. (default: `minimize_depth`) |
-| `decomposition_type` | Choose one of the following: (a) `exact_optimal`, which finds an exact, provably optimal, decomposition if it exists, (c) `exact_feasible`, which finds any feasible exact decomposition, but not necessarily an optimal one if it exsists (c) `approximate`, which finds an approximate decomposition if an exact one does not exist; otherwise it will return an exact decomposition (default: `exact_optimal`)|
+| `decomposition_type` | Choose one of the following: (a) `exact_optimal`, which finds an exact, provably optimal, decomposition if it exists, (b) `exact_feasible`, which finds any feasible exact decomposition, but not necessarily an optimal one if it exsist, (c) `exact_optimal_global_phase`, which finds an optimal circuit decomposition if it exists, up to a global phase (d) `approximate`, which finds an approximate decomposition if an exact one does not exist; otherwise it will return an exact decomposition (default: `exact_optimal`)|


Could you please correct exsist to exists.

including this in the new commit

harshangrjn · 2022-10-05T17:18:06Z

src/data.jl

+            ref_nonzero_r, ref_nonzero_c = QCO._get_ref_nonzero_index_of_original_target(params["target_gate"])
+            global_phase = angle(params["target_gate"][ref_nonzero_r, ref_nonzero_c])
+            is_target_real_up_to_phase = QCO.is_gate_real(exp(-im*global_phase)*params["target_gate"])
+            if is_target_real_up_to_phase
+                return real(exp(-im*global_phase)*params["target_gate"]), !is_target_real


Could this also be written as a general evaluation for global_phase = 0 and then evaluate the global phase of target if exact_optimal_global_phase is true? Also, we should add a test to cover this case where target is real in a global_phase sense.

implemented the global_phase 0, added tests and including this in the new commit

harshangrjn · 2022-10-05T17:20:27Z

src/log.jl

+    global_phase = 1
+
+    if data["decomposition_type"] in ["exact_optimal_global_phase"]
+        n_r = size(target_gate)[1]


Could these refs be directly obtained using your helper function in utility.jl?

changed so this uses: _get_ref_nonzero_index_of_original_target

harshangrjn · 2022-10-05T17:25:08Z

src/constraints.jl

+function constraint_gate_target_condition_compact_global_phase_real(qcm::QuantumCircuitModel)
+
+    max_depth   = qcm.data["maximum_depth"]
+    n_r            = size(qcm.data["gates_real"])[1]
+    n_c            = size(qcm.data["gates_real"])[2]
+
+    U_var = qcm.variables[:U_var]
+
+    ref_nonzero_r, ref_nonzero_c = QCO._get_ref_nonzero_index_of_real_target(qcm.data::Dict{String,Any})
+
+    for i=1:n_r, j=1:n_c
+        if isapprox(qcm.data["target_gate"][i,j], 0, atol=1E-6)
+            JuMP.@constraint(qcm.model, U_var[i,j,max_depth] == 0)
+        else
+            JuMP.@constraint(qcm.model, U_var[i,j,max_depth]*qcm.data["target_gate"][ref_nonzero_r,ref_nonzero_c] == qcm.data["target_gate"][i,j]*U_var[ref_nonzero_r,ref_nonzero_c,max_depth])
+        end
+    end
+    return    


This function needs to be tested in unit tests. Also, I think this and constraint_gate_target_condition_compact_global_phase_complex could be combined into a single function?

combined the two global_phase_constraints
unit test now covering both cases

…l phase constraints

harshangrjn · 2022-10-07T23:22:18Z

Thanks @szabino-mrszabo

szabino-mrszabo added 22 commits September 21, 2022 09:37

test commit

3e83a9a

implementig global phase

d80115e

fixing validation validate_circuit_decomposition for global phase

e6fcd38

experimenting with toffoli

c56bf52

redefining toffoli

07abbf4

if all gates real but target is a phase times real, ignore the phase …

f6b475d

…and set target as the phaseless target

implementing simple chain constrain

bcf0082

backrolling the change to chain constrain

190c0c4

adding 7q cat state verification code

688fdea

adding decompose q5 random target

f6d8b2a

more parameter changes

f09b15f

changing nonzero to nonzero_ref

579be65

implementing U[depth] n exact_optimal_global_phase

ff79a9d

Merge branch 'master' of https://github.com/szabino-mrszabo/szabo-Qua…

8423d01

…ntumCircuitOpt.jl

getting harsha changes

bdd2575

Merge v0.5.0 into global_phase

9f11811

fixing last merge issues from gl_p - V0.5.0

6e652ce

adding test for global_phase

bd77780

add variable

8eb03d5

back to previous

963ce78

working gl phase

c24df4c

ne need for complex constraint for U[max_depth]

8074405

harshangrjn self-requested a review October 4, 2022 23:53

harshangrjn reviewed Oct 5, 2022

View reviewed changes

harshangrjn added this to the v0.5.1 milestone Oct 6, 2022

addressing comments, adding more unit tests for QC_model Tests: Globa…

53ff494

…l phase constraints

harshangrjn merged commit cb8ee64 into harshangrjn:master Oct 7, 2022

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

New feature: Global phase linear constraints #62

New feature: Global phase linear constraints #62

szabino-mrszabo commented Oct 4, 2022

codecov bot commented Oct 4, 2022 •

edited

Loading

harshangrjn left a comment

harshangrjn Oct 5, 2022

szabino-mrszabo Oct 5, 2022

harshangrjn Oct 5, 2022

szabino-mrszabo Oct 6, 2022

harshangrjn Oct 5, 2022

szabino-mrszabo Oct 5, 2022

harshangrjn Oct 5, 2022

szabino-mrszabo Oct 5, 2022

harshangrjn Oct 5, 2022

szabino-mrszabo Oct 6, 2022

harshangrjn Oct 5, 2022

szabino-mrszabo Oct 5, 2022

harshangrjn commented Oct 7, 2022

New feature: Global phase linear constraints #62

New feature: Global phase linear constraints #62

Conversation

szabino-mrszabo commented Oct 4, 2022

codecov bot commented Oct 4, 2022 • edited Loading

Codecov Report

harshangrjn 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

harshangrjn commented Oct 7, 2022

codecov bot commented Oct 4, 2022 •

edited

Loading