Wednesday, December 14, 2022
Time: 16:10-17:00
venue: Mathematics Building Room 527
Prof. Cheng-Fang Su
Department of Applied Mathematics, National Yang Ming Chiao Tung University
Title:
Two Research Results in Quantum Machine Learning
Abstract
This talk covers two quantum machine-learning topics. In topic one, we introduce the method to classify data on tumor metastasis effectively by support vector algorithms. A few studies suggest that quantum support vector machine algorithms perform well in classification problems. Furthermore, if doctors can identify biomarkers to predict tumor metastasis accurately, it will be an essential step toward precision medicine. Therefore, we use both the SVM and QSVM classifiers, adding a certain number of features and evaluating the performance of quantum and classical algorithms in classifying tumor metastasis data. Finally, we achieve excellent distinctions between patients with or without metastasis. This is joint work with Tai-Yue Li, Venugopala Reddy Mekala, and Ka-Lok Ng.
Topic two, previous research has demonstrated that classical classifiers are vulnerable to adversarial examples, and their risk of subjecting to adversarial attack surges exponentially with even small increments in the data dimension. So we want to overcome noise in the quantum computing process, as it will exacerbate the process or render the result useless. However, our research argues that these two diverging directions can be combined to create an intriguing property. Namely, we discover that utilizing the added quantum random rotation noise, which simulates the natural noise source in the quantum computing process, can improve the robustness of quantum classifiers against adversarial attacks. Moreover, based on the property of differential privacy of quantum classifiers, we derive a certified robustness bound to defend against all general adversaries. This is joint work with Jhih- Cing Huang, Yu-Lin Tsai, Chao-Han Huck Yang, Chia-Mu Yu, Pin-Yu Chen, and Sy-Yen Kuo.