Quantum approximate optimization algorithm
Background
The Quantum Approximate Optimization Algorithm (QAOA) is a hybrid iterative method for solving combinatorial optimization problems. In this tutorial, we demonstrate how to implement the QAOA algorithm using Qiskit Runtime for solving a simple max-cut problem.
In a max-cut problem, we want to partition nodes of a graph in a way that maximizes the number of edges between nodes in differing groups. The desired max-cut partition for the following graph is clear: the 0th-node on the left should be separated from the rest of the nodes on the right by a cut. We will find this answer by applying QAOA by using Qiskit Runtime primitives and sessions.
Setup
No output produced
Output:
'ibmq_mumbai'
Step 1: Map classical inputs to a quantum problem
To demonstrate max-cut, we'll create a graph using the rustworkx library, and create Pauli Hamiltonian that encodes the cost in a manner such that the minimum expectation value of the operator corresponds to the maximum number of edges between the nodes in two different groups.
Output:
For this simple example, the operator is a linear combination of terms with
Output:
The previous image illustrates the ansatz in basic gates for clarity. However, it can be expressed in multiple levels of decomposition by changing the
Output:
Step 2: Optimize problem for quantum execution.
To reduce the total job execution time, V2 primitives only accept circuits (ansatz) and observables (Hamiltonian) that conforms to the instructions and connectivity supported by the target system (referred to as instruction set architecture (ISA) circuits and observables).
ISA Circuit
We can schedule a series of
optimization_level : The lowest optimization level just does the bare minimum needed to get the circuit running on the device; it maps the circuit qubits to the device qubits and adds swap gates to allow all 2-qubit operations. The highest optimization level is much smarter and uses lots of tricks to reduce the overall gate count. Since multi-qubit gates have high error rates and qubits decohere over time, the shorter circuits should give better results.
No output produced
Output:
ISA Observables
Similarly, we need to transform the Hamiltonian to make it backend compatible before running jobs with
Output:
SparsePauliOp(['IIIIIIIIIIIIZIIIIIZIIIIIIII', 'IIIIIIIIIIIIZIIZIIIIIIIIIII', 'IIIIIIIIIIIIZZIIIIIIIIIIIII', 'IIIIIIIIIIZIZIIIIIIIIIIIIII'],
coeffs=[1.+0.j, 1.+0.j, 1.+0.j, 1.+0.j])
Step 3: Execute using Qiskit Primitives.
As with an iterative optimization procedure, we now need to define our cost function over which to minimize. We proceed in an identical manner to the Variational Quantum Eigensolver tutorial, computing the expectation value of our Hamiltonian with respect to the parameterized ansatz circuit using the Qiskit Runtime
No output produced
Any classical optimizer can be used to minimize the cost function. On a real quantum system, an optimizer designed for non-smooth cost function landscapes usually does better. Here we use the COBYLA routine from SciPy through the minimize function.
Because we are iteratively executing many calls to Runtime, we use a
No output produced
We can set an initial set of random parameters:
No output produced
We can run our minimization routine:
No output produced
In the end, we have a result in the standard SciPy
Output:
message: Optimization terminated successfully.
success: True
status: 1
fun: -2.482839158666561
x: [ 5.054e-01 5.090e+00 2.655e+00 5.840e+00]
nfev: 47
maxcv: 0.0
Step 4: Post-process, return result in classical format.
The solution vector of parameter angles (
Output:
No output produced
For small problem instances, the solution can be visually obtained:
Output:
The most probable bit-strings, up to finite-sampling deviations, encode the solution. Here we see that
Output:
Output:
'0.21.1'
Output:
'1.0.1'
Was this page helpful?