Quantum Circuit Mapping via Reinforcement Learning and MCTS

Solution Overview

Problem

Quantum computing technologies face challenges in efficiently mapping logical quantum circuits to physical hardware devices due to limitations in qubit connectivity, leading to increased execution time and error rates from SWAP operations, which are necessary to circumvent physical limitations but introduce noise and complexity.

Innovation Solution

The use of reinforcement learning techniques, specifically a policy-based neural network assisted by a Monte Carlo Tree Search algorithm, to optimize quantum circuit mapping by minimizing the number of SWAP operations and efficiently allocate and route quantum gates on available qubits, thereby addressing the NP-hard optimization problem.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If SWAP operations are used to circumvent physical qubit connectivity limitations, then quantum circuits can be executed on hardware with restricted connectivity, but execution time increases and error rates increase due to noise from additional operations

Engineering Contradiction:

Improvequbit connectivity compatibilityVSAvoidexecution error rate

Core Design Contradiction:

Adaptability or versatilityVSReliability

Solution Approach 1:

The reinforcement learning model performs preliminary planning of the entire quantum circuit execution before actual execution, predicting optimal qubit allocations and SWAP operation sequences in advance. This allows the system to minimize unnecessary SWAP operations while ensuring connectivity constraints are met, thereby reducing execution time and error rates

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system uses reinforcement learning with reward functions that provide feedback during the mapping process. The reward signal guides the model to select qubit allocations and SWAP operations that minimize execution time and error rates while satisfying connectivity constraints, enabling continuous optimization of the circuit mapping

Inventive Principle:

Principle #23Feedback

2Adaptability or versatility

If SWAP operations are used to circumvent physical qubit connectivity limitations, then quantum circuits can be executed on hardware with restricted connectivity, but execution time increases due to additional operations

Engineering Contradiction:

Improvequbit connectivity compatibilityVSAvoidcircuit execution time

Core Design Contradiction:

Adaptability or versatilityVSDuration of action of moving object

Solution Approach 1:

The reinforcement learning model performs preliminary planning of the entire quantum circuit execution before actual execution, predicting optimal qubit allocations and SWAP operation sequences in advance. This allows the system to minimize unnecessary SWAP operations while ensuring connectivity constraints are met, thereby reducing execution time

Inventive Principle:

Principle #10Preliminary action

Solution Approach 2:

The system dynamically adjusts mapping parameters such as qubit allocation strategies and SWAP operation timing based on the specific circuit structure and hardware connectivity. By optimizing these parameters through reinforcement learning, the system reduces the total number of operations and minimizes execution time

Inventive Principle:

Principle #35Parameter changes

3Productivity

If reinforcement learning with Monte Carlo Tree Search is used to optimize quantum circuit mapping, then the number of SWAP operations is minimized and routing is optimized, but computational resources and complexity increase

Engineering Contradiction:

Improvecircuit mapping efficiencyVSAvoidoptimization system complexity

Core Design Contradiction:

ProductivityVSDevice complexity

Solution Approach 1:

The optimization process is segmented into distinct phases: Monte Carlo Tree Search for exploring possible mappings, reinforcement learning for optimizing selected mappings, and compilation for final circuit generation. This segmentation allows each component to be optimized independently and facilitates parallel processing, reducing overall computational burden

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The system applies partial optimization by focusing computational resources on the most critical parts of the circuit mapping, such as identifying bottleneck operations and optimizing high-impact qubit allocations. This selective approach achieves significant improvements without requiring exhaustive optimization of all circuit elements

Inventive Principle:

Principle #16Partial or excessive action

Data Source

PatentUS20240330730A1Quantum circuit mapping using reinforcement learning techniques

Publication Date: 2024.10.03 AMAZON TECH INC

AI summary

Techniques for solving quantum circuit mapping problems using reinforcement learning techniques are disclosed. Quantum circuit mapping often requires the use of SWAP gates in order to configure logical quantum computations to be executed using fixed quantum hardware device layouts. A reinforcement learning model takes inputs such as a logical quantum circuit, a physical qubit connectivity graph corresponding to a quantum hardware device, and an initial qubit allocation scheme, and uses such information to schedule quantum gates of the logical quantum circuit for execution using respective physical qubits of the quantum hardware device. A reinforcement learning model that is configured to solve such quantum circuit mapping problems may comprise a neural network that is assisted by a Monte Carlo Tree Search (MCTS) algorithm, wherein the MCTS algorithm guides the neural network towards quantum circuit routing pathways which are more efficient (e.g., require fewer SWAP gates to be scheduled).