Min Chen

Min Chen's primary research interest is data science in general and data visualization in particular, and he considers Data Science as the scientific discipline that studies human and machine processes for transforming data to decisions and/or knowledge. Its main goal is to understand the inner workings of different data intelligence processes, such as statistical inference, algorithmic reasoning, human thinking, and collaborative decision making, and to provide a scientific foundation to underpin the design, engineering, and optimization of data intelligence workflows composed of human and machine processes. Theoretical Data Science is a major branch of Data Science that focuses on the mathematical theories that underpin all aspects of data science and enable abstract modelling of data intelligence workflows. Applied Data Science is a major branch of Data Science that focuses on the technologies (e.g., data mining, data visualization, machine learning, etc.) for supporting the design, engineering and optimization of data intelligence processes and workflows. Min Chen has made technical contributions to the following research topics: Visual Analytics. For many data intelligence problems, there is no fully-automated solution, likely because of the complex information space, the inadequate sampling (e.g. sparse or dated samples), the complexity of the algorithm, and so on. Visual analytics represents a methodology for bringing machine-centric processes (e.g., statistics and algorithms) and human-centric processes (e.g., visualization and interaction) together and designing an optimised workflow for such data intelligence problems. In this respect, the scientific essence is to understand and measure the relative merits of machine and human processes, while the engineering essence is optimisation rather than brute-force automation. Theories, Metrics and Empirical Studies. In the scientific world, a theory is a fact-based framework for explaining a set of observed phenomena or events. It is typically formulated to facilitate falsifiable predictions about some causal relations. It often involves quantitative measures, enabling analytical inference and numerical simulation. At the moment, information theory, which underpins tele- and data communication, has shown to be able to explain numerous phenomena of perception, cognition, emotion, and interaction in visualization. The key to such an explanation is the counter-intuitive cost-benefit measure, where entropy reduction (or information loss) is viewed as a merit rather than a demerit. The cost-benefit measure has been successfully used to analyse the effectiveness of performing visualization tasks in different virtual environments, optimise the trade-off between human and machine processes in data intelligence workflows, quantify informative contributions of human knowledge to machine learning, and explain the role of human knowledge in visualization processes, especially when there is significant information loss (e.g., underground maps and volume visualization). Video Visualization. Video visualization is concerned with the creation of a new visual representation from an input video to reveal important features and events in the video. It typically extracts meaningful information from a video and conveys the extracted information to users in abstract or summary visual representations, which are typically more compact than the input video itself. Video visualization is not intended to provide fully automatic solutions to the problem of making decisions about the contents of a video. Instead, it aims at offering a tool to assist users in their intelligent reasoning while removing or reducing the burden of viewing videos. In particular, it can be used to fill in many gaps in practice where automated computer vision is yet to provide deployable solutions. This aim justifies deviation from the creation of realistic imagery, and allows simplifications and embellishments, to improve the understanding of the input video. The fundamental challenge is: Can we see time (i.e., temporal information) without using time (i.e., an animation)? Volume Graphics and Volume Visualization. Volume graphics is concerned with graphics scenes, where models are defined using volume representations instead of, or in addition to, traditional surface representations. It is a study of the input, storage, construction, manipulation, display, and animation of volume models in a true three-dimensional (3D) form. Its primary aim is to create realistic and artistic computer-generated imagery from graphics scenes comprising volume objects, and to facilitate the interaction with these objects in graphical virtual environments. Volume visualization is also concerned with volume data representations that are used to store measured physical attributes of real-world objects and phenomena, or to represent computer-generated models and their attributes in volumetric forms. Although it is typical and conventional for volume datasets (such as in computed tomography) to correspond spatially to the 3D physical world, it is also common in some visualization applications to use volume datasets to store non-spatial physical data as well as abstract information.

[2023] Measurement of the azimuthal anisotropy of charged particles in sNN=5.36TeV O16+O16 and Ne20+Ne20 collisions with the ATLAS detector [2023] AEM: An interpretable multi-task multi-modal framework for cardiac disease prediction [2023] Dimensionality Reduction with Entropies from f-Divergences [2023] An Anatomical Significance-Aware Architecture for Explainable Myocardial Infarction Prediction via Multi-task Learning [2023] How to Reject a VIS Paper, or Not?