arXiv:2101.03181 · Machine Learning Science and Technology
Miniaturizing neural networks for charge state autotuning in quantum dots
Stefanie Czischek, Victor Yon, Marc-Antoine Genest, Marc-Antoine Roux, Sophie Rochette, Julien Camirand Lemyre, Mathieu Moras, Michel Pioro-Ladrière, Dominique Drouin, Yann Beilliard, Roger G. Melko
Abstract A key challenge in scaling quantum computers is the calibration and control of multiple qubits. In solid-state quantum dots (QDs), the gate voltages required to stabilize quantized charges are unique for each individual qubit, resulting in a high-dimensional control parameter space that must be tuned automatically. Machine learning techniques are capable of processing high-dimensional data—provided that an appropriate training set is available—and have been successfully used for autotuning in the past. In this paper, we develop extremely small feed-forward neural networks that can be used to detect charge-state transitions in QD stability diagrams. We demonstrate that these neural networks can be trained on synthetic data produced by computer simulations, and robustly transferred to the task of tuning an experimental device into a desired charge state. The neural networks required for this task are sufficiently small as to enable an implementation in existing memristor crossbar arrays in the near future. This opens up the possibility of miniaturizing powerful control elements on low-power hardware, a significant step towards on-chip autotuning in future QD computers.
Abstract A key challenge in scaling quantum computers is the calibration and control of multiple qubits. In solid-state quantum dots (QDs), the gate voltages required to stabilize quantized charges are unique for each individual qubit, resulting in a high-dimensional control parameter space that must be tuned automatically. Machine learning techniques are capable of processing high-dimensional data—provided that an appropriate training set is available—and have been successfully used for autotuning in the past. In this paper, we develop extremely small feed-forward neural networks that can be used to detect charge-state transitions in QD stability diagrams. We demonstrate that these neural networks can be trained on synthetic data produced by computer simulations, and robustly transferred to the task of tuning an experimental device into a desired charge state. The neural networks required for this task are sufficiently small as to enable an implementation in existing memristor crossbar arrays in the near future. This opens up the possibility of miniaturizing powerful control elements on low-power hardware, a significant step towards on-chip autotuning in future QD computers.
Spin-based semiconductor qubits hold promise for scalable quantum computing, yet they require reliable autonomous calibration procedures. This study presents an experimental demonstration of online single-dot charge autotuning using a convolutional neural network integrated into a closed-loop calibration system. The autotuning algorithm explores the gates' voltage space to localize charge transition lines, thereby isolating the one-electron regime without human intervention. This exploration leverages the model's uncertainty estimation to find the appropriate gate configuration with minimal measurements while reducing the risk of failures. In 20 experimental runs, our method achieved a success rate of 95% in locating the target electron regime, highlighting the robustness of this approach against noise and distribution shifts from the offline training set. Each tuning run lasted an average of 2 h and 9 min, primarily due to the limited speed of the current measurement. This work validates the feasibility of machine-learning-driven real-time charge autotuning for quantum dot devices, advancing the development toward the control of large qubit arrays.
Abstract This study presents a machine learning-based procedure to automate the charge tuning of semiconductor spin qubits with minimal human intervention, addressing one of the significant challenges in scaling up quantum dot technologies. This method exploits artificial neural networks to identify noisy transition lines in stability diagrams, guiding a robust exploration strategy leveraging neural network uncertainty estimations. Tested across three distinct offline experimental datasets representing different single-quantum-dot technologies, this approach achieves a tuning success rate of over 99% in optimal cases, where more than 10% of the success is directly attributable to uncertainty exploitation. The challenging constraints of small training sets containing high diagram-to-diagram variability allowed us to evaluate the capabilities and limits of the proposed procedure.
Abstract Recent progress in building large-scale quantum devices for exploring quantum computing and simulation has relied upon effective tools for achieving and maintaining good experimental parameters, i.e., tuning up devices. In many cases, including quantum dot-based architectures, the parameter space grows substantially with the number of qubits, and may become a limit to scalability. Fortunately, machine learning techniques for pattern recognition and image classification, using so-called deep neural networks, have shown surprising successes for computer-aided understanding of complex systems. We propose a new paradigm for fully automated experimental initialization through a closed-loop system relying on machine learning and optimization techniques. We use deep convolutional neural networks to characterize states and charge configurations of semiconductor quantum dot arrays when only measurements of a current−voltage characteristic of transport are available. For simplicity, we model a semiconductor nanowire connected to leads and capacitively coupled to depletion gates using the Thomas−Fermi approximation and Coulomb blockade physics. We then generate labeled training data for the neural networks, and find at least 90 % accuracy for charge and state identification for single and double dots. Using these characterization networks, we can then optimize the parameter space to achieve a desired configuration of the array, a technique we call “auto-tuning”. Finally, we show how such techniques can be implemented in an experimental setting by applying our approach to an experimental dataset, and outline further problems in this domain, from using charge sensing data to extensions to full one- and two-dimensional arrays, that can be tackled with machine learning.
Abstract Tuning of gate-defined semiconductor quantum dots (QDs) is a major bottleneck for scaling spin-qubit technologies. We present a deep learning driven, semantic-segmentation pipeline that performs charge auto-tuning by locating transition lines in full charge stability diagrams (CSDs) and returns gate voltage targets for the single charge regime. We assemble and manually annotate a large, heterogeneous dataset of 1015 experimental CSDs measured from fully depleted silicon-on-insulator (FD-SOI) QD devices, spanning nine design geometries, multiple wafers, and fabrication runs. A U-Net style convolutional neural network with a MobileNetV2 encoder is trained and validated through five-fold group cross-validation. Our model achieves an overall offline tuning success of 80.0% in locating the single-charge regime, with peak performance exceeding 88% for some designs. We analyze dominant failure modes and propose targeted mitigations. Finally, wide-range diagram segmentation also naturally enables scalable physics-based feature extraction that can feed back to fabrication and design workflows and outline a roadmap for real-time integration in a cryogenic wafer prober. Overall, our results show that neural network based wide-diagram segmentation is a practical step toward automated, high-throughput charge tuning for silicon QD qubits.
The development of energy-efficient and fast machine learning methods plays an increasingly important role in experimental physics, where data analysis and control tasks often need to operate under strict resource constraints. In these contexts, machine learning models can automate complex calibration and analysis tasks while enabling on-device data processing close to the experimental sensors.One representative application presented on this poster concerns the automated calibration of semiconductor spin qubits, while the outlook highlights extensions toward edge-computing approaches in detector systems. The automated calibration of quantum dots is a key prerequisite for realizing scalable quantum computers. In particular, the analysis of charge stability diagrams, used to detect charge transitions in quantum dots, represents a complex and time-consuming task. Neural networks, especially U-Net architectures, offer the potential to automate this process by reliably recognizing relevant patterns in simulated and experimental measurement data. State-of-the-art networks have already been successfully trained for this purpose. However, there remains significant potential for optimization to enable space- and energy-efficient integration close to the quantum bits within the cryostat. We have investigated the use of quantized neural networks for energy-efficient quantum dot calibration. The goal is to analyze the impact of post-training quantization and quantization-aware training on detection quality, as well as the general effects of quantization on memory requirements and inference speed. Three U-Nets with different architectures, parameter counts, and input dimensions serve as model bases, applied to simulated charge stability diagrams. The results show that appropriate quantization strategies can reduce memory usage without significantly affecting detection quality. The findings of this work contribute to the integration of energy-efficient machine learning methods into experimental quantum computing environments, thereby supporting overall scalability. Building on these results, the approach is extended toward the use of binarized neural networks (BNNs) to push energy efficiency and faster inference even further. Within edge computing applications, current efforts focus on implementing and demonstrating such networks on FPGA hardware, aiming to exploit binary-weight computation and hardware-level parallelism for minimal latency and power consumption. Beyond quantum dot calibration, these methods are also being investigated for other scientific applications, such as the autonomous self-triggering radio detection of extensive air showers, highlighting the broader potential of hardware-embedded AI for resource-constrained experimental environments.
The development of energy-efficient and fast machine learning methods plays an increasingly important role in experimental physics, where data analysis and control tasks often need to operate under strict resource constraints. In these contexts, machine learning models can automate complex calibration and analysis tasks while enabling on-device data processing close to the experimental sensors.One representative application presented on this poster concerns the automated calibration of semiconductor spin qubits, while the outlook highlights extensions toward edge-computing approaches in detector systems.The automated calibration of quantum dots is a key prerequisite for realizing scalable quantum computers. In particular, the analysis of charge stability diagrams, used to detect charge transitions in quantum dots, represents a complex and time-consuming task. Neural networks, especially U-Net architectures, offer the potential to automate this process by reliably recognizing relevant patterns in simulated and experimental measurement data. State-of-the-art networks have already been successfully trained for this purpose.However, there remains significant potential for optimization to enable space- and energy-efficient integration close to the quantum bits within the cryostat.We have investigated the use of quantized neural networks for energy-efficient quantum dot calibration. The goal is to analyze the impact of post-training quantization and quantization-aware training on detection quality, as well as the general effects of quantization on memory requirements and inference speed. Three U-Nets with different architectures, parameter counts, and input dimensions serve as model bases, applied to simulated charge stability diagrams. The results show that appropriate quantization strategies can reduce memory usage without significantly affecting detection quality.The findings of this work contribute to the integration of energy-efficient machine learning methods into experimental quantum computing environments, thereby supporting overall scalability.Building on these results, the approach is extended toward the use of binarized neural networks (BNNs) to push energy efficiency and faster inference even further. Within edge computing applications, current efforts focus on implementing and demonstrating such networks on FPGA hardware, aiming to exploit binary-weight computation and hardware-level parallelism for minimal latency and power consumption. Beyond quantum dot calibration, these methods are also being investigated for other scientific applications, such as the autonomous self-triggering radio detection of extensive air showers, highlighting the broader potential of hardware-embedded AI for resource-constrained experimental environments.