Quantum Unsupervised Learning for Anti-Money Laundering Detection

机遇：金融犯罪是风险管理中增长最快的领域之一。洗钱尤其对国家和国际安全构成重大威胁。侦查金融犯罪的方法不足会给金融机构和社会带来严重后果。金融服务机构及其监管机构认为，开发改进的检测和减少洗钱的方法势在必行。**方法：**量子算法在经典数据上的异常检测研究还很不足。该项目将增强现有的量子机器学习模型，超越监督学习，以检测表明洗钱活动的异常行为。在这项工作中，我们将利用推广投影量子核方法\[1\]的技术，将量子计算机的输出作为经典机器学习算法的组成部分。该项目将利用Rigetti的量子计算机，其可扩展的超导芯片架构，支持快速、高保真的纠缠门。Rigetti最近解锁了快速执行投影量子核及其泛化的可能性。**创新和好处：**机器学习技术通过标记潜在的可疑交易和适应不断变化的犯罪行为，在检测和预防金融犯罪方面具有很大的前景。量子计算可以作为经典计算的加速器。通过将现有的经典机器学习技术扩展到量子计算增强能力，金融机构将更好地应对金融犯罪。开发改进的机器学习能力来打击金融犯罪将为金融机构提供竞争优势，加速量子计算的商业化，并提供更有效的识别和减少犯罪活动的方法。通过加入汇丰银行深厚的领域专业知识，爱丁堡大学的算法专业知识，Rigetti最先进的量子计算硬件和软件，以及国家量子计算中心在英国量子计算生态系统中的关键作用，该项目不仅将利用行业领先的资源和世界级的研究人员，但它将为未来的量子算法和应用开发铺平道路，使金融服务行业和许多其他垂直行业受益。\[1\] Huang等人，“量子机器学习中数据的力量”，2021年2月，[https://arxiv.org/abs/2011.01938][0].[0]: https://arxiv.org/abs/2011.01938]

**The Opportunity:** Financial crime is one of the fastest-growing areas of risk management. Money laundering in particular poses a significant threat to national and international security. Inadequate methods for detecting financial crimes has serious consequences for financial organisations and society. Developing improved methods of detecting and reducing money laundering is seen as imperative by financial services institutions and their regulators. **The Approach:** Quantum algorithms for anomaly detection on classical data are rather understudied. This project will enhance current quantum machine learning models beyond supervised learning to detect anomalous behaviour indicating money laundering activity. In this work, we will use a quantum computer's output as a component to a classical machine learning algorithm by leveraging techniques that generalise projected quantum kernel methods \[1\]. The project will utilise Rigetti's quantum computer, with its scalable superconducting chip architecture, supporting rapid, high fidelity entangling gates. Rigetti has recently unlocked the possibility of executing projected quantum kernels and their generalisations at speed. **Innovation and Benefits:** Machine learning technology holds great promise to detect and prevent financial crime by flagging potential suspicious transactions and adapting to ever-changing criminal behaviour. Quantum computing can serve as an accelerator to classical computing. By extending capabilities from existing classical machine learning techniques to quantum computing enhanced capabilities, financial institutions will be better equipped to tackle financial crime. Developing improved machine learning capabilities to combat financial crime will provide financial institutions a competitive advantage, accelerate the commercialisation of quantum computing, and provide more effective ways of identifying and reducing criminal activity. By joining HSBC's deep domain expertise, the University of Edinburgh's algorithm expertise, Rigetti's state-of-the-art quantum computing hardware and software, and the National Quantum Computing Centre's pivotal role in the UK's quantum computing ecosystem, this project will not only harness industry-leading resources and world-class researchers, but will pave the way for future quantum algorithm and application development to benefit the financial services industry and numerous other verticals.\[1\] Huang et al., "Power of data in quantum machine learning," February 2021, [https://arxiv.org/abs/2011.01938][0].[0]: https://arxiv.org/abs/2011.01938