Research Approach and Interests
The Machine Learning for Signal Processing and Communications Group (MLSC Group) designs numerical optimization algorithms to solve different machine/deep learning problems and apply the solution to different applications, such as data association for automotive radar, channel estimation and precoder design for wireless communications, data analysis using distributed and federated learning, and graph signal processing and graph learning for communications and biomedical engineering. We leverage our knowledge in a signal processing (SP) and numerical optimization during the design process which allows us to extend conventional model-based SP approaches to model/data-driven or pure data-driven algorithms.
My current research is mainly focused divided into three parts: 1) Graph signal processing (GSP) and graph learning, 2) distributed and federated learning, 3) 6G communications.
<![if !supportLists]>· <![endif]>GSP and graph learning
<![if !supportLists]>o <![endif]>We have been focusing on online graph learning which tracks on time-varying graph that models non-Euclidean data that lie on an irregular structure. Applications include modeling of cyberattacks, brain and network traffic. Graph filter, graph neural networks and graph convolutional neural networks can be used in conjunction with the learned graph to solve community detection problem related to cyberattacks, prediction of onset of brain disease, and transceiver design in wireless communication systems.
<![if !supportLists]>· <![endif]>Distributed and federated learning (edge computing) in Heterogeneous Networks
<![if !supportLists]>o <![endif]>We have been looking at the problem of supervised, self-supervised, and unsupervised distributed and federated learning, with the first two aiming to train MLPs and model-based deep learning models in heterogeneous networks (with statistical and system heterogeneity)
<![if !supportLists]>o <![endif]>We will be looking how to improve learning of non-Euclidean using kernel learning techniques
<![if !supportLists]>o <![endif]>Distributed online graph learning problem becomes essential to prevent raw data from being transmitted to the cloud for learning task
<![if !supportLists]>· <![endif]>Signal processing for 6G communications
<![if !supportLists]>o <![endif]>This work has mainly been focused on channel estimation and precoder design in intelligent reflective surface-assisted (IRS-assisted) systems, using (convex) optimal design methodology and model-based deep learning, with extension to using federated learning framework.
<![if !supportLists]>o <![endif]>We plan to investigate IRS-assisted system operating at mid-band frequency of 7 – 24 GHz, which has not been assigned by the 3GPP committee, but is able to support fast data speed (compared to sub-6 GHz) and reasonable coverage