[Paper Review] Reconstructing charged particle track segments with a quantum-enhanced support vector machine

Philippa Kathryn Duckett, G. Facini+4 more|arXiv (Cornell University)|Dec 14, 2022

Particle Detector Development and Performance36 references4 citations

TL;DR

This paper proposes a hybrid quantum-classical support vector machine (SVM) with a quantum-estimated kernel to classify charged particle track segments in high-luminosity LHC detector data. It demonstrates improved accuracy over classical SVMs for triplets from the innermost detector layers—critical for seeding track reconstruction—marking the first application of quantum-kernel SVMs to particle track reconstruction.

ABSTRACT

Reconstructing the trajectories of charged particles from the collection of hits they leave in the detectors of collider experiments like those at the Large Hadron Collider (LHC) is a challenging combinatorics problem and computationally intensive. The ten-fold increase in the delivered luminosity at the upgraded High Luminosity LHC will result in a very densely populated detector environment. The time taken by conventional techniques for reconstructing particle tracks scales worse than quadratically with track density. Accurately and efficiently assigning the collection of hits left in the tracking detector to the correct particle will be a computational bottleneck and has motivated studying possible alternative approaches. This paper presents a quantum-enhanced machine learning algorithm that uses a support vector machine (SVM) with a quantum-estimated kernel to classify a set of three hits (triplets) as either belonging to or not belonging to the same particle track. The performance of the algorithm is then compared to a fully classical SVM. The quantum algorithm shows an improvement in accuracy versus the classical algorithm for the innermost layers of the detector that are expected to be important for the initial seeding step of track reconstruction.

Motivation & Objective

• Address the computational bottleneck in charged particle track reconstruction at the upgraded High Luminosity LHC (HL-LHC), where track density and pileup are expected to increase dramatically.
• Explore quantum-enhanced machine learning as a potential solution to accelerate and improve the accuracy of track reconstruction in high-density detector environments.
• Investigate whether quantum-estimated kernels in SVMs can outperform classical kernels in classifying track segment triplets from simulated detector hits.
• Focus on the seeding phase of track reconstruction, where accuracy in the innermost detector layers is most critical for downstream performance.

Proposed method

• Use a publicly available TrackML dataset simulating HL-LHC detector conditions to extract doublets and triplets of consecutive hits from reconstructed tracks.
• Apply a hybrid quantum-classical SVM framework where the kernel matrix is estimated using a parameterized quantum circuit (PQC) on a NISQ-era quantum device.
• Train the quantum-estimated kernel SVM on nine-feature triplets representing spatial coordinates and geometric properties of three consecutive hits.
• Compare performance against a classical SVM using a standard RBF kernel and a baseline using randomly selected triplets from the dataset.
• Evaluate classification performance using accuracy scores, with a focus on spatial regions of the detector.
• Use quantum kernel estimation techniques that allow for approximate kernel computation via quantum circuits, with error bounds analyzed for potential speedup.

Experimental results

Research questions

• RQ1 Can a quantum-estimated kernel in an SVM outperform a classical kernel in classifying track segment triplets in high-density particle detector data?
• RQ2 Does the quantum-enhanced approach show measurable improvement in accuracy for the innermost layers of the tracking detector, which are most critical for seeding track reconstruction?
• RQ3 What is the impact of quantum noise on the reliability and precision of kernel estimation in this quantum-classical machine learning pipeline?
• RQ4 Under what conditions—such as dataset size, feature count, or qubit count—might quantum kernel estimation offer a computational advantage over classical methods?
• RQ5 How does the performance of the quantum kernel SVM scale with increasing complexity of the track segment classification task?

Key findings

• The quantum-estimated kernel SVM achieves a similar level of overall accuracy compared to the classical SVM across the full dataset.
• For triplets originating from the innermost layers of the tracking detector, the quantum algorithm shows a measurable improvement in accuracy over the classical SVM.
• The improvement in accuracy for inner-layer triplets is significant because these hits are most critical for the seeding step in track reconstruction algorithms.
• The results suggest that quantum kernel estimation may be particularly advantageous in high-feature, low-to-moderate dataset regimes, such as those involving geometric track segment classification.
• The study identifies potential for future quantum advantage in track reconstruction, especially if kernel estimation methods are optimized to reduce noise sensitivity and improve scalability.
• Noise in current NISQ devices causes kernel entries to concentrate around a fixed value, requiring exponentially many measurement shots to resolve differences, which poses a challenge for practical implementation.