Artificial Intelligence and Machine Learning (AIML)
Research in the AIML Group aims to create a unique synergy of strength in artificial intelligence, machine learning, big data analytics, data mining, intelligent user interfaces, autonomy-oriented computing, and Web intelligence. The group is positioned to develop advanced algorithms and intelligent systems to meet the real-world needs and challenges, e.g., in healthcare and social computing.
Faculty Involved:
- Prof. CHEN, Li
- Dr. CHEN, Yifan
- Prof. CHEUNG, William Kwok Wai
- Dr. HAN, Bo
- Prof. LIU, Jiming
- Dr. LIU, Yang
- Dr. QI, Jun
- Dr. YIN, Kejing
- Dr. ZHANG, Eric Lu
- Dr. ZHU, Ziqing
Funded Research and Consultancy Projects in the Past Few Years:
| How Serendipity Improves User Satisfaction with Recommendations? A Large-Scale User Evaluation |
 | Staff |
Dr. CHEN, Li |
| Abstract |
- To identify the causal relationships from novelty, unexpectedness, relevance, and timeliness to serendipity, and from serendipity to user satisfaction and purchase intention.
- To reveal the moderating effects of user curiosity on the relationships from novelty to serendipity and from serendipity to satisfaction.
- To compare four algorithms for recommending e-commerce products in terms of user perceptions.
|
| Macro environment and non-company news on stock market performance: A comparative study of media sentiment, public engagement, and prediction models |
 | Staff |
Dr. CHEN, Li |
| Abstract |
- To analyze the news coverage by Hong Kong media in past 20 years, and illustrate the longitudinal patterns of news sentiment and related public engagement on interactive media outlets.
- To explore and discover several useful non-company specific and macro environment indicators from the news patterns that are significantly correlated with market (Hang Seng Index and sub-indices) and specific stock price fluctuations.More
- To compare predictability and reliability of the found non-company specific news indicators and that of existing company specific tools in financial practices.
- To develop accurate and efficient computational methods for market indicators identification and media text analysis.
|
| User Evaluation on Sentiment-based Recommendation Explanations |
 | Staff |
Dr. CHEN, Li |
| Abstract |
- To investigate how review information could be exploited to generate explanations for multiple recommendations especially in high-investment product domains (i.e., digital camera and laptop).
- To identify the exact effect of sentiment-based explanations on improving users’ decision effectiveness and system perceptions.More
- To investigate how users view information and compare products on the sentiment-based explanation interface.
|
| Detecting the Tipping Points from Inflammation to Tumorigenesis Using A Multi-level Network Modeling Approach |
 | Staff |
Dr. CHEUNG, William Kwok Wai |
| Abstract |
- To detect tipping points by identifying “dynamical network biomarker” at molecular network level by analysing the ‘omics’ data.
- To identify potential networks associated with tipping points by constructing multi-level regulatory networks from heterogeneous data sources and performing structural analysis on evolving networksMore
- To conduct biological experiments for validating our findings of potential networks associated with the tipping points.
|
| Joint Embedding of Sequential Data and Knowledge Graphs with application to Predictive Analytics in Healthcare |
 | Staff |
Dr. CHEUNG, William Kwok Wai |
| Abstract |
- To develop methodologies that can learn representations simultaneously from heterogeneous clinical events in the EHR data and the medical knowledge graphs for patient outcome prediction with enhanced interpretability.
- To develop an interpretable framework to jointly infer the correspondence among different modalities in the heterogeneous EHR data and generate high-throughput phenotype candidates.
|
| Trustworthy Deep Learning from Open-set Corrupted Data |
 | Staff |
Dr. HAN, Bo |
| Abstract |
- Developing a dual-scored methodology to model open-set instance-dependent noisy labels robustly; Designing instance-level learning algorithms with theoretical guarantees to solve the proposed model.
- Exploiting generalized unlabelled data as auxiliary medium to robustly handle open-set adversarial examples; Leveraging adversarial robust loss to jointly train on original training set and unlabelled data with pseudo-labels.
More
- Designing an adversarial dual checking methodology to robustly adapt from corrupted source domain to open-set unlabelled target domain.
- Automating and integrating above orthogonal techniques into an Automated Trustworthy Deep Learning (AutoTDL) system; Testing this system using real-world corrupted data.
|
| Learning Discriminative Manifolds from Multiple Data Modalities for Multimedia Content Understanding |
 | Staff |
Dr. LIU, Yang |
| Abstract |
To discover the intrinsic structure of multimedia data with multiple modalities for multimedia content understanding
- For multi-label multimedia data, learn the correlation between different labels while preserving the discriminative information.
- For multi-modal multimedia data, learn the consistency between different modalities while preserving the discriminative
information.
|
