Quantum Circuit Architecture Screening for Variational ML Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Designing efficient quantum gate sequences for variational quantum circuits is challenging due to the computational expense of simulating larger numbers of qubits and quantum gates on classical hardware, making trial and error methods inconclusive.
Innovation Solution
A computer-implemented method for constructing quantum circuit-based machine learning models by sampling random quantum circuits, evaluating them with circuit metrics like Fourier expressivity, ZX-calculus redundancy, and Fisher information, and optimizing variational parameters to determine suitable quantum circuit architectures.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If quantum circuits are simulated on classical hardware to evaluate different quantum circuit architecture options, then circuit design can be performed and evaluated, but the simulation becomes computationally expensive for larger numbers of qubits
Solution Approach 1:
The patent applies preliminary action by pre-evaluating quantum circuit architectures using simulation before deployment on actual quantum hardware. Multiple circuit configurations are simulated in advance to identify promising candidates, so that when deployed on real quantum devices, the most suitable architectures are already selected, avoiding expensive trial-and-error on actual hardware.
Solution Approach 2:
The patent uses copying by creating and evaluating multiple simulated copies of quantum circuits on classical hardware. Instead of testing each circuit configuration on physical quantum devices, virtual copies are simulated repeatedly to assess performance, enabling efficient comparison of many architectural options without consuming quantum hardware resources.
2Adaptability or versatility
If trial and error methods are used to design quantum circuit architectures, then different configurations can be explored, but the methods become inconclusive for larger numbers of qubits due to computational expense
Solution Approach 1:
The patent implements feedback by systematically evaluating simulated quantum circuits against performance metrics and using the results to guide subsequent design iterations. The simulation results provide feedback on which architectural features perform well, allowing the design process to converge on optimal configurations rather than relying on random trial and error.
Solution Approach 2:
The patent applies parameter changes by systematically varying quantum circuit architecture parameters (such as gate types, circuit depth, qubit connectivity patterns) in simulations to identify which parameter combinations yield optimal performance. This structured parameter exploration is more efficient than unguided trial and error because it uses simulation feedback to guide parameter selection.
3Quantity of substance
If a large number of qubits and quantum gates are used in variational quantum circuits to provide quantum advantage, then access to large internal state space is achieved, but simulation on classical hardware becomes intractable
Solution Approach 1:
The patent applies segmentation by dividing the quantum circuit design process into multiple stages: initial exploration with smaller circuits, intermediate validation, and final optimization. This segmentation allows systematic scaling to larger qubit counts by progressively building complexity rather than attempting to simulate all configurations at full scale simultaneously.
Solution Approach 2:
The patent uses partial action by simulating and evaluating only subsets of the full quantum circuit configuration space. Instead of exhaustively simulating all possible circuits with large numbers of qubits, the method selectively evaluates representative samples and uses those results to infer performance of similar configurations, reducing simulation burden while maintaining design quality.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A computer-implemented method for constructing a quantum circuit-based machine learning model to estimate an output for a set of input features based on problem data, comprising receiving a quantum circuit architecture selection including a selection of a quantum circuit architecture parameter range for specifying a property of a quantum circuit layer, sampling random quantum circuits based on the quantum circuit architecture selection with a random selection of quantum circuit architecture parameters from the quantum circuit architecture parameter range, determining a circuit metric for each of the random quantum circuits, determining, based on the circuit metric meeting a corresponding metric criterion, a subset of random quantum circuits as validated quantum circuit layer candidates, training machine learning models, each comprising one of the validated quantum circuit based layer candidates; and determining an optimized set of quantum circuit architecture parameters based on a quality metric achieved by the machine learning models.