21 The noise control of the sound field in the cabin is very important to the comfort of passengers, and to a large extent depends on the successful numerical simulation of the high-frequency sound field. The time-harmonic sound field in a car cabin is typically described by the solution of the Helmholtz equation. The numerical solution of the Helmholtz equation with large wave number is still a computational challenge nowadays. Prof. Buyang LI and his team developed a new cutting-edge computational method for solving the high-frequency Helmholtz equation in three-dimensional large-scale computations. The resulting partial differential equation solver outperforms existing commercial software and is being incorporated into the industrial automobile manufacturing in Huawei. High-frequency Simulation Technology of Car Cockpit Sound Field AMA conducts high-impact research with applications in engineering, science, communication, finance and retailing. Department of Applied Mathematics Solving Large-scale Linear Programming Models for Production Planning Production planning is an essential part of the daily operations of Huawei, where huge-scale linear programming problems need to be solved. Our team developed a series of novel techniques including the sGS-ADMM warm-up tool for solving these problems. The resulting optimisation solver outperforms existing commercial software and has been incorporated into the daily production planning in Huawei. In recognition of the contributions made by our research team, Prof. Defeng SUN has been awarded the Distinguished Collaborator Award both from Hong Kong Research Center and Noah’s Ark Lab, Huawei Technologies Co. Ltd. for this collaborative project. Multi-layer Segmentation of Retina OCT Images Optical Coherence Tomography (OCT) is a non-invasive method which can obtain high-definition images of cross-section (B-scan) of the retina. By investigating the thickness of different layers of the retina in OCT images, one can diagnose ocular diseases in an early stage. We have developed several AI techniques based on U-net like structures to train models for the prediction of retinal layers. In reducing the complexity of networks, a method is proposed based on the concept of domain decomposition when training a large volume of data on a cloud platform. Prof. Cedric YIU’s team is currently collaborating with practitioners in Ophthalmology for the potential use in OCT machines. Our Research Collaborations with Industry
RkJQdWJsaXNoZXIy Mjc5OTU=