ckchui

In the past 10+ years, laser microscopy has successfully made it permeated not only in biochemistry and cell/molecular biology research, but also in numerous preclinical and clinical applications. However, our understanding of health and disease is still very limited. This lecture will introduce the latest breakthrough in laser microscopy technologies developed at HKU that can address some of these challenges. Especially these technologies can generate unprecedented views and understanding of the living biological cells. They include: capturing high-resolution motion picture of the swift-flying brain signals in a living animal; visualizing the inner workings of biological cells and organisms in 3D without killing them; and detecting rare cancer cells in millions of blood cells. Not only can these technologies impact new biological discovery (e.g. neuroscience), but also creating many new opportunities in cost-effective clinical diagnosis, especially cancer screening.

In the past 10+ years, laser microscopy has successfully made it permeated not only in biochemistry and cell/molecular biology research, but also in numerous preclinical and clinical applications. However, our understanding of health and disease is still very limited. This lecture will introduce the latest breakthrough in laser microscopy technologies developed at HKU that can address some of these challenges. Especially these technologies can generate unprecedented views and understanding of the living biological cells. They include: capturing high-resolution motion picture of the swift-flying brain signals in a living animal; visualizing the inner workings of biological cells and organisms in 3D without killing them; and detecting rare cancer cells in millions of blood cells. Not only can these technologies impact new biological discovery (e.g. neuroscience), but also creating many new opportunities in cost-effective clinical diagnosis, especially cancer screening.

The diamond has been well known as the gem stones in jewellery market, and the same material with various atomic defects, i.e., fluorescent impurities in diamond lattice, shows unique quantum behaviors even at ambient conditions. A diamond, not just a best friend of ladies, but also the best friend of scientists. Due to their unique quantum properties, these atomic defects has been demonstrated to achieve nanometric measurement of various physical quantities such as electromagnetic fields, temperature and etc. with unprecedented precision. Here, I will firstly review the development of diamond-based science and technology, and discuss its potential applications in diverse fields. Specifically, I will introduce the on-going research activities in my group, mainly including the high figure-of-merit diamond materials synthesis, advanced quantum diamond microscope development and diamond quantum sensing in single living cells. In addition, I will also share my journey in exploring beyond academics, e.g., we apply quantum diamond microscope for authenticity identification in local jewellery industry.

Stainless steel (SS) is one of the most extensively used materials in many public areas and hygiene facilities but has no inherent antimicrobial properties. Additionally, the SARS-CoV-2 exhibits strong stability on regular SS surfaces, with viable viruses detected even after three days. Undoubtedly, this has created a high possibility of virus transmission among people using these areas and facilities. Here, this talk presents the inactivation of pathogen microbes (especially the SARS-CoV-2) on SS surface by tuning the chemical composition and microstructure of regular SS. It is discovered that Pathogen viruses like H1N1 and SARS-CoV-2 exhibit good stability on the surface of pure Ag and Cu-contained SS of low Cu content, but are rapidly inactivated on the surface of pure Cu and Cu-contained SS of high Cu content. Significantly, the developed anti-pathogen SS with 20 wt% Cu can distinctly reduce 99.75% and 99.99% of viable SARS-CoV-2 on its surface within 3 and 6 h, respectively. Lift buttons made of the present anti-pathogen SS are produced using mature powder metallurgy technique, demonstrating its potential applications in public areas and fighting the transmission of SARS-CoV-2 and other pathogens via surface touching.

Membrane separation technology is increasingly used for water and energy related applications. Pressure-driven membrane processes, such as microfiltration (MF), ultrafiltration (UF), nanofiltration (NF), and reverse osmosis (RO), have received great attention, fueled by the increasing needs for water purification, wastewater treatment and reclamation, and seawater desalination. In parallel, many novel membranes and membrane processes are being developed. In this TechTalk, Prof. Chuyang Tang will share his personal journey in the amazing membrane world. He will highlight some of his previous and ongoing research works covering topics on water reuse, seawater desalination, resource recovery, energy production, and beyond.

Distributed training using multiple devices (i.e., GPU servers) has been widely adopted for learning DNN models over large datasets. However, the performance of large-scale distributed training tends to be far from linear speed-up in practice. Given the complexity of distributed systems, it is challenging to identify the root cause(s) of inefficiency and exercise effective performance optimizations when unexpected low training speed occurs. To date, there exists no software tool which diagnoses performance issues and helps expedite distributed DNN training, while the training can be run using different machine learning frameworks. This paper proposes dPRO, a toolkit that includes: (1) an efficient profiler that collects runtime traces of distributed DNN training across multiple frameworks, especially fine-grained communication traces, and constructs global data flow graphs including detailed communication operations for accurate replay; (2) an optimizer that effectively identifies performance bottlenecks and explores optimization strategies (from computation, communication and memory aspects) for training acceleration. We implement dPRO on multiple deep learning frameworks (PyTorch, TensorFlow, MXNet) and representative communication schemes (AllReduce and Parameter Server architecture). Extensive experiments show that dPRO predicts performance of distributed training in various settings with<5% errors in most cases and finds optimization strategies with up to87.1%speed-up over the baselines.