In the Centre for Industrial Software (CIS) the research area of Quantum Machine Learning (QML) represents an intersection of quantum computing and machine learning, leveraging quantum mechanics to enhance the efficiency and capabilities of traditional machine learning algorithms.
Quantum bits, or qubits, store information in quantum computing and, unlike classical bits, can exist in superposition—a state where they simultaneously represent 0 and 1. This property allows quantum computers to process multiple computations in parallel, vastly increasing computational power for certain tasks. In QML, algorithms utilize the procedures of entanglement and superposition to simultaneously explore multiple solutions, thereby improving their search capabilities when dealing with complex datasets. This process operates within the framework of Hilbert space, a fundamental concept in quantum mechanics that provides a rich mathematical structure for representing and manipulating quantum states. QML can handle large amounts of data and do operations that aren't possible with classical systems because it uses Hilbert space. This has made huge strides in solving difficult problems in many areas of interest.
Our QML activities in the CIS aim to create methods and tools suitable for a wider range of applications, including IoT, CPS, robotics, and drones, and explore two complementary pathways:
Distributed & resilient machine learning
We propose a novel multimodal quantum federated learning framework that utilizes quantum computing to counteract the performance drop resulting from Fedrared HE. For the first time in FL, our framework combines a multimodal quantum mixture of experts (MQMoE) model with FHE, incorporating multimodal datasets for enriched representation and task-specific learning. Our MQMoE framework enhances performance on multimodal datasets and combined genomics and brain MRI scans, especially for underrepresented categories. Our results also demonstrate that the quantum-enhanced approach mitigates the performance degradation associated with FHE and improves classification accuracy across diverse datasets, validating the potential of quantum interventions in enhancing privacy in FL.
Vulnerabilities recognition
While QML presents promising avenues for advancing cybersecurity capabilities, it is imperative to acknowledge and address the inherent limitations and challenges of its application.
- Find effective ways to deal with model biases, resource limitations in quantum computing, and implementation difficulties if we want to get the most out of quantum-enhanced cybersecurity solutions
- Resource optimization should be a top priority. Enhancing the performance of quantum computing and minimizing its workload can be achieved by designing efficient quantum circuits and effectively managing qubits. This involves developing algorithms, such as noise-adaptive search, that maximize the efficient use of available qubits and quantum gates. By creating robust quantum circuits, we ensure that quantum computations are both effective and resource-efficient.
- Enhanced explainability in QML arises from the rich representational capabilities of quantum systems. This could make models more accurate and interpretable using algorithms like quantum Shapley values. QML's ability to learn and represent complex data structures more effectively enhances its explainability. This is particularly important in cybersecurity, where understanding the reasoning behind model decisions is crucial for trust and accountability.
Contact
See more
QADQN: Quantum Attention Deep Q-Network for Financial Market Prediction