Researchers from the Korea Advanced Institute of Science and Technology and LG Electronics Toronto AI Lab are developing quantum circuits to refine how measurements are taken on quantum hardware. Their work compares three distinct approaches, Naimark measurements, hybrid Naimark-QNN measurements and fully QNN measurements, for strategies like state discrimination. The team demonstrates that quantum neural network circuits can achieve near-optimal quantum measurements with fewer training iterations, a key advantage for quantum machine learning. This research presents constructions of quantum circuits to implement general measurements on quantum hardware.
Naimark Extension Constructs Measurement Circuits via Unitary Transformations
The Naimark extension provides a method for constructing quantum circuits capable of implementing general measurements on quantum hardware, using unitary transformations and ancillary qubits. This approach allows any positive operator-valued measure, or POVM, to be realized as a projective measurement within an expanded Hilbert space. Specifically, the researchers constructed a circuit, termed a “measurement circuit,” to build a Kraus operator, where the circuit operates on both the system and ancilla qubits to determine measurement probabilities.
The mathematical foundation rests on the Naimark dilation theorem, which enables embedding a quantum system into a larger space for projective measurement. A key aspect of this work involves framing a circuit ansatz based on the Naimark extension, using universal gate sets like controlled-NOT and single-qubit gates. These single-qubit gates are initially defined with parameters, then refined using a classical optimizer to approximate a desired quantum measurement.
The complexity of implementing a POVM directly correlates to the construction of this quantum circuit. The researchers point out that while current quantum technologies are suited for shallow-depth circuits, achieving full control over quantum errors remains a challenge for polynomial-depth implementations. The resulting POVM, constructed through the Naimark extension, is termed a “Naimark measurement,” and precisely constructs a desired POVM using elementary gates. “An arbitrary measurement described by a POVM can be implemented by a measurement circuit via the Naimark extension,” the paper states.
The work demonstrates that a quantum circuit can implement a general measurement via the Naimark extension, performing a circuit over system and ancilla qubits to implement a general operation described by a Kraus operator. The construction of a circuit U(SA), over system and ancilla qubits, is central to this process, with outcomes on the ancilla qubits defining the measurement.
The study details how the circuit U(SA) constructs a Kraus operator with fixed ancillary qubits prepared in the computational basis, as described in Equation 1. This approach allows for the implementation of l-outcome POVM elements, M = Mi(i = 1)l, through the Naimark extension, ensuring that the sum of all measurement outcomes equals the identity operator.
Hybrid Naimark-QNN Circuits Reduce Complexity with Parameterized Quantum Circuits
By framing quantum measurements within the Naimark extension and using parameterized circuits, scientists have created a pathway to reduce computational demands during optimization. This contrasts with the computational difficulty posed by the ZZ-interactions inherent in traditional Naimark measurement circuits. Replacing these interaction blocks with quantum neural network circuits, specifically parameterized circuits, forms the basis of hybrid Naimark-QNN measurements, significantly reducing the number of CNOT gates required.
The researchers also explored fully QNN measurement circuits, constructed solely from these neural network building blocks, dispensing with the Naimark extension altogether. This exploration allowed for a direct comparison of measurement circuit designs: Naimark, hybrid Naimark-QNN and fully QNN, tested against state discrimination strategies like minimum-error and maximum-confidence measurements.
QNN Measurements Achieve Near-Optimal State Discrimination with Fewer Iterations
This construction allows for direct comparison against Naimark and hybrid Naimark-QNN designs in the task of quantum state discrimination, specifically using both minimum-error and maximum-confidence strategies. The work demonstrates that QNN circuits approximate Naimark measurements efficiently, while also offering a computational benefit; a classical optimizer converges more rapidly when applied to QNN measurements than to their Naimark counterparts. Specifically, the researchers found that QNN measurements achieve comparable performance with a reduced number of training iterations, a critical factor in accelerating quantum machine learning workflows.
The efficiency stems from a restructuring of the underlying circuit architecture. Traditional Naimark measurements often incorporate two-qubit gates involving ZZ-interactions, which present computational challenges. By replacing these interaction blocks with parameterized circuits, the hybrid and fully QNN designs reduce complexity and facilitate optimization.
The team also developed a block-encoding circuit specifically to realize maximum-confidence measurements, further expanding the versatility of the QNN approach. “QNN measurement circuits are near-optimal in that they efficiently approximate Naimark measurements,” the paper states, highlighting the balance between accuracy and computational cost. The researchers’ findings suggest a pathway toward more streamlined and effective quantum measurement techniques, essential for advancing quantum information processing.




See today’s quantum computing news on Quantum Zeitgeist for the latest breakthroughs in qubits, hardware, algorithms, and industry deals.
