Merge pull request #61 from harshangrjn/U_matrix_fixed_idx

Fixed unitary matrix elements

CHANGELOG.md

@@ -1,11 +1,21 @@
 QuantumCircuitOpt.jl Change Log
 ===============================
 
+### v0.5.0
+- Minor: Fixed result `primal_status` issue in `log.jl`
+- Added helper functions for obtaining fixed indices of `U_var` unitary variables to zeros or constant values. Dropped linearization constraints for fixed `U_var` variables
+- Dropped support for `unit_magnitude_constraints`
+- Reformulated quadratic objective function of `slack_var` variables into a linear outer-approximation - increases code coverage due to MILP reformulation 
+- Fixed `slack_var` based on fixed `U_var` variables  
+- Minor updates in error messages for kron symboled gates
+- Changed Cbc test dependency to HiGHS (MIP) solver 
+- Updated docs and unit tests to reflect above changes
+
 ### v0.4.1
 - Added support for returning exact feasible decomposition (without optimality certificate) using `exact_feasible`
 - Added support for unitary constraints using binary linearizations (speeds up feasiblity problems)
-- Added `decompose_W_using_HSTCnot` in examples
-- Added `decompose_CNOT_using_GR` in examples (for Rydberg atom array-based simulators)
+- Added `W_using_HSTCnot` in examples
+- Added `CNOT_using_GR` in examples (for Rydberg atom array-based simulators)
 - Added support for `CSGate`, `CSdaggerGate`, `CTGate`, `CTdaggerGate` and `SSwapGate` (sqrt(`SwapGate`))
 - Dropped support for `CZRevGate` as it is invariant to qubit flip
 - Updated README and docs with the Youtube link for the JuliaCon's presentation

Project.toml

@@ -1,7 +1,7 @@
 name = "QuantumCircuitOpt"
 uuid = "88128e30-b60a-4e54-ab02-1050a5f92a36"
 authors = ["Harsha Nagarajan"]
-version = "0.4.1"
+version = "0.5.0"
 
 [deps]
 DataStructures = "864edb3b-99cc-5e75-8d2d-829cb0a9cfe8"
@@ -16,8 +16,8 @@ Memento = "~1.0, ~1.1, 1"
 julia = "^1"
 
 [extras]
-Cbc = "9961bab8-2fa3-5c5a-9d89-47fab24efd76"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+HiGHS = "87dc4568-4c63-4d18-b0c0-bb2238e4078b"
 
 [targets]
-test = ["Cbc", "Test"]
+test = ["HiGHS", "Test"]

README.md

@@ -5,6 +5,7 @@
 Status: 
 [![CI](https://github.com/harshangrjn/QuantumCircuitOpt.jl/actions/workflows/ci.yml/badge.svg)](https://github.com/harshangrjn/QuantumCircuitOpt.jl/actions/workflows/ci.yml)
 [![codecov](https://codecov.io/gh/harshangrjn/QuantumCircuitOpt.jl/branch/master/graph/badge.svg?token=KGJWIV6QF4)](https://codecov.io/gh/harshangrjn/QuantumCircuitOpt.jl)
+[![version][https://juliahub.com/docs/QuantumCircuitOpt/version.svg]][https://juliahub.com/ui/Packages/QuantumCircuitOpt/dwSy1]
 
 <!-- Stable version: [![Documentation](https://github.com/harshangrjn/QuantumCircuitOpt.jl/actions/workflows/documentation.yml/badge.svg)](https://harshangrjn.github.io/QuantumCircuitOpt.jl/stable/) -->
 Dev version: [![Documentation](https://github.com/harshangrjn/QuantumCircuitOpt.jl/actions/workflows/documentation.yml/badge.svg)](https://harshangrjn.github.io/QuantumCircuitOpt.jl/dev/)

docs/src/index.md

@@ -22,11 +22,11 @@ import Pkg
 Pkg.add("QuantumCircuitOpt")
 ```
 
-At least one mixed-integer programming solver is required for running QuantumCircuitOpt. The well-known [CPLEX](https://github.com/jump-dev/CPLEX.jl) or the [Gurobi](https://github.com/jump-dev/Gurobi.jl) solver is highly recommended, as it is fast, scaleable and can be used to solve on fairly large-scale graphs. However, open-source solvers such as [Cbc](https://github.com/jump-dev/Cbc.jl) or [GLPK](https://github.com/jump-dev/GLPK.jl) is also compatible with QuantumCircuitOpt which can be installed via the package manager with
+At least one mixed-integer programming (MIP) solver is required for running QuantumCircuitOpt. The well-known [Gurobi](https://github.com/jump-dev/Gurobi.jl) or IBM's [CPLEX](https://github.com/jump-dev/CPLEX.jl) solver is highly recommended, as it is fast, scaleable and can be used to solve on fairly large-scale circuits. However, the open-source MIP solver [HiGHS](https://github.com/jump-dev/HiGHS.jl) is also compatible with QuantumCircuitOpt. Gurobi (or any other MIP solver) can be installed via the package manager with
 
 ```julia
 import Pkg
-Pkg.add("Cbc")
+Pkg.add("Gurobi")
 ```
 
 ## Unit Tests

docs/src/quickguide.md

@@ -34,7 +34,7 @@ Building on the recent success of [Julia](https://julialang.org), [JuMP](https:/
 ```
 
 ## Getting started
-To get started, install [QCOpt](https://github.com/harshangrjn/QuantumCircuitOpt.jl) and [JuMP](https://github.com/jump-dev/JuMP.jl), a modeling language layer for optimization. QCOpt also needs a MIP solver such as [CPLEX](https://github.com/jump-dev/CPLEX.jl) or [Gurobi](https://github.com/jump-dev/Gurobi.jl). If you prefer an open-source MIP solver, install [CBC](https://github.com/jump-dev/Cbc.jl) or [GLPK](https://github.com/jump-dev/GLPK.jl) from the Julia package manager, though be warned that the run times of QCOpt can be substantially slower using these open-source MIP solvers. 
+To get started, install [QCOpt](https://github.com/harshangrjn/QuantumCircuitOpt.jl) and [JuMP](https://github.com/jump-dev/JuMP.jl), a modeling language layer for optimization. QCOpt also needs a MIP solver such as [Gurobi](https://github.com/jump-dev/Gurobi.jl) or IBM's [CPLEX](https://github.com/jump-dev/CPLEX.jl). If you prefer an open-source MIP solver, install [HiGHS](https://github.com/jump-dev/HiGHS.jl) from the Julia package manager, though be warned that the run times of QCOpt can be substantially slower using any of the open-source MIP solvers. 
 
 # User inputs
 QCOpt takes two types of user-defined input specifications. The first type of input contains all the necessary circuit specifications. This is given by a dictionary in Julia, which is a collection of key-value pairs, where every key is of the type `String`, which admits values of various types. Below is the list of allowable keys for the dictionary, given in column 1, and it's respective values with descriptions, given in column 2. This input dictionary is represented as `params` in all the [example](https://github.com/harshangrjn/QuantumCircuitOpt.jl/tree/master/examples) circuit decompositions. 
@@ -88,7 +88,7 @@ The second set of inputs for QCOpt contains all the optional specifications for
 |`identity_gate_symmetry_constraints`| This option activates the valid constraints to eliminate symmetry in the Identity gate in the decomposition (default: `true`)| 
 |`idempotent_gate_constraints`| This option activates the valid constraints to eliminate idempotent gates in the elementary (native) gates set (default: `false`)| 
 |`convex_hull_gate_constraints`| This option activates the valid constraints to apply convex hull of complex entries in the elementary (native) gates set (default: `false`)| 
-|`unit_magnitude_constraints`| This option activates the valid outer-approximation constraints for unit-valued complex entries of the unitary gates (`U_var`) (default: `false`)|
+|`fix_unitary_variables`| This option evaluates all the fixed-valued indices of unitary matrix varaibles (`U_var`) at every depth, and appropriately builds the optimization model (default: `true`)|
 |`visualize_solution`| This option activates the visualization of the optimal circuit decomposition (default: `true`)| 
 | `relax_integrality` | This option transforms integer variables into continuous variables (default: `false`).  |
 | `optimizer_log` | This option enables or disables console logging for the `optimizer` (default: `true`).|
@@ -161,23 +161,25 @@ QCOpt.visualize_solution(results, data)
 For example, for the above controlled-Z gate decomposition, the processed output of QCOpt is as follows: 
 ```
 =============================================================================
+QuantumCircuitOpt version: v0.5.0
+
 Quantum Circuit Model Data
 
   Number of qubits: 2
   Total number of elementary gates (after presolve): 72
   Maximum depth of decomposition: 4
-  Input elementary gates: ["U3_1", "U3_2", "CNot_1_2", "Identity"]
+  Elementary gates: ["U3_1", "U3_2", "CNot_1_2", "Identity"]
     U3_θ discretization: [-180.0, -90.0, 0.0, 90.0, 180.0]
     U3_ϕ discretization: [-180.0, -90.0, 0.0, 90.0, 180.0]
     U3_λ discretization: [-180.0, -90.0, 0.0, 90.0, 180.0]
-  Type of decomposition: exact
+  Type of decomposition: exact_optimal
   MIP optimizer: Gurobi
 
 Optimal Circuit Decomposition
 
   U3_2(-90.0,0.0,0.0) * CNot_1_2 * U3_2(90.0,0.0,0.0) = Target gate
   Minimum optimal depth: 3
-  Optimizer run time: 2.64 sec.
+  Optimizer run time: 3.01 sec.
 =============================================================================
 ```
 

examples/2qubit_gates.jl

@@ -1,4 +1,4 @@
-function decompose_hadamard()
+function hadamard()
 
     println(">>>>> Hadamard Gate <<<<<")
 
@@ -17,7 +17,7 @@ function decompose_hadamard()
     )
 end
 
-function decompose_controlled_Z()
+function controlled_Z()
 
     println(">>>>> Controlled-Z Gate <<<<<")
 
@@ -37,7 +37,7 @@ function decompose_controlled_Z()
 
 end
 
-function decompose_controlled_V()
+function controlled_V()
 
     println(">>>>> Controlled-V Gate <<<<<")
 
@@ -52,7 +52,7 @@ function decompose_controlled_V()
     )
 end
 
-function decompose_controlled_H()
+function controlled_H()
 
     println(">>>>> Controlled-H Gate <<<<<")
 
@@ -71,7 +71,7 @@ function decompose_controlled_H()
     )
 end
 
-function decompose_controlled_H_with_R()
+function controlled_H_with_R()
 
     println(">>>>> Controlled-H with R Gate <<<<<")
 
@@ -90,7 +90,7 @@ function decompose_controlled_H_with_R()
     )
 end
 
-function decompose_magic()
+function magic()
 
     println(">>>>> M Gate <<<<<")
 
@@ -109,7 +109,7 @@ function decompose_magic()
         )
 end
 
-function decompose_magic_using_CNOT_1_2()
+function magic_using_CNOT_1_2()
 
     println(">>>>> M Gate <<<<<")
 
@@ -128,7 +128,7 @@ function decompose_magic_using_CNOT_1_2()
         )
 end
 
-function decompose_magic_using_SHCnot()
+function magic_using_SHCnot()
 
     println(">>>>> M gate using S, H and CNOT Gate <<<<<")
 
@@ -143,7 +143,7 @@ function decompose_magic_using_SHCnot()
         )
 end
 
-function decompose_S()
+function S_gate()
 
     println(">>>>> S Gate <<<<<")
 
@@ -162,7 +162,7 @@ function decompose_S()
         )
 end
 
-function decompose_revcnot()
+function revcnot()
 
     println(">>>>> Reverse CNOT Gate <<<<<")
 
@@ -173,11 +173,11 @@ function decompose_revcnot()
     "elementary_gates" => ["H_1", "H_2", "CNot_1_2", "Identity"],  
     "target_gate" => QCOpt.CNotRevGate(),
     "objective" => "minimize_depth", 
-    "decomposition_type" => "exact_optimal"     
+    "decomposition_type" => "exact_optimal"
     )
 end
 
-function decompose_revcnot_with_U()
+function revcnot_with_U()
 
     println(">>>>> Reverse CNOT using U3 and CNOT Gates <<<<<")
 
@@ -196,7 +196,7 @@ function decompose_revcnot_with_U()
         )
 end
 
-function decompose_swap()
+function swap()
 
     println(">>>>> SWAP Gate <<<<<")
 
@@ -211,7 +211,7 @@ function decompose_swap()
         )
 end
 
-function decompose_W()
+function W_gate()
 
     println(">>>>> W Gate <<<<<")
 
@@ -230,7 +230,7 @@ function decompose_W()
         )
 end
 
-function decompose_W_using_HSTCnot()
+function W_using_HSTCnot()
 
     println(">>>>> W Gate using H, S, T and CNOT gates <<<<<")
 
@@ -245,7 +245,7 @@ function decompose_W_using_HSTCnot()
     )
 end
 
-function decompose_W_using_HCnot()
+function W_using_HCnot()
 
     println(">>>>> W Gate using H and CNOT gates <<<<<")
 
@@ -260,7 +260,7 @@ function decompose_W_using_HCnot()
         )
 end
 
-function decompose_HCoinGate()
+function HCoinGate()
 
     println(">>>>> Hadamard Coin gate <<<<<")
 
@@ -275,7 +275,7 @@ function decompose_HCoinGate()
         )
 end
 
-function decompose_GroverDiffusion_using_HX()
+function GroverDiffusion_using_HX()
 
     println(">>>>> Grover's Diffusion Operator <<<<<")
 
@@ -292,7 +292,7 @@ function decompose_GroverDiffusion_using_HX()
         )
 end
 
-function decompose_GroverDiffusion_using_Clifford()
+function GroverDiffusion_using_Clifford()
 
     println(">>>>> Grover's Diffusion Operator <<<<<")
 
@@ -308,7 +308,7 @@ function decompose_GroverDiffusion_using_Clifford()
 end
 
 
-function decompose_GroverDiffusion_using_U3()
+function GroverDiffusion_using_U3()
 
     println(">>>>> Grover's Diffusion Operator using U3 gate <<<<<")
 
@@ -327,7 +327,7 @@ function decompose_GroverDiffusion_using_U3()
     )
 end
 
-function decompose_iSwap()
+function iSwap()
 
     println(">>>>> iSwap Gate <<<<<")
 
@@ -339,11 +339,11 @@ function decompose_iSwap()
         # "elementary_gates" => ["S_1", "S_2", "Sdagger_1", "Sdagger_2", "H_1", "H_2", "CNot_1_2", "CNot_2_1", "Identity"],
         "target_gate" => QCOpt.iSwapGate(),
         "objective" => "minimize_depth",
-        "decomposition_type" => "exact_optimal",                           
+        "decomposition_type" => "exact_optimal",  
         )
 end
 
-function decompose_qft2_using_R()
+function qft2_using_R()
     println(">>>>> QFT2 Gate using R and CNOT gates <<<<<")
 
     return Dict{String, Any}(
@@ -353,14 +353,14 @@ function decompose_qft2_using_R()
         "elementary_gates" => ["RX_1", "RZ_1", "RZ_2", "CNot_1_2", "CNot_2_1", "Identity"],
         "target_gate" => QCOpt.QFT2Gate(),
         "objective" => "minimize_depth",
-        "decomposition_type" => "approximate",
+        "decomposition_type" => "exact_feasible",
 
         "RX_discretization" => [π/2],
         "RZ_discretization" => [-π/4, π/2, 3*π/4, 7*π/4],                      
         )
 end
 
-function decompose_qft2_using_HT()
+function qft2_using_HT()
     println(">>>>> QFT2 Gate using H, T, CNOT gates <<<<<")
 
     return Dict{String, Any}(
@@ -374,11 +374,11 @@ function decompose_qft2_using_HT()
         )
 end
 
-function decompose_GRGate()
+function GRGate()
     println(">>>>> GRGate testing <<<<<")
 
     GR1 = QCOpt.get_full_sized_gate("GR", 2; angle = [π/6, π/3])
-    GR2 = QCOpt.get_full_sized_gate("GR", 2; angle = [π/3, π/6])
+    # GR2 = QCOpt.get_full_sized_gate("GR", 2; angle = [π/3, π/6])
     CNot_1_2 = QCOpt.get_full_sized_gate("CNot_1_2", 2)
     T = QCOpt.round_complex_values(GR1 * CNot_1_2)
 
@@ -395,7 +395,7 @@ function decompose_GRGate()
         )
 end
 
-function decompose_X_using_GR()
+function X_using_GR()
 
     println(">>>>> Pauli-X Gate using global rotation <<<<<")
 
@@ -415,7 +415,7 @@ function decompose_X_using_GR()
     )
 end
 
-function decompose_CNOT_using_GR()
+function CNOT_using_GR()
     println(">>>>> CNOT Gate using global rotation <<<<<")
 
     return Dict{String, Any}(