Lab Leader
- Nayyar Zaidi DMKD Group Lead
DMKD Group
The DMKD Group works on data mining, knowledge discovery, and machine learning methods for structured, multimodal, scientific, and engineering data. Our research focuses on building models that are not only accurate, but also useful under practical constraints: limited data, heterogeneous data, domain knowledge, robustness, uncertainty, and deployment.
Photos
Midyear group retreat
Group photo after research discussions and project updates.
Research discussion session
Students and collaborators presenting emerging ideas and ongoing work.
Christmas retreat
End-of-year gathering to celebrate progress, publications, and collaborations.
Group Lunch
Group Lunch at Deakin Downtown.
KDD2026 Oral Talk
Presentation at KDD2026 Jeju, Korea.
ICRA Poster Presentation
Presentation at ICRA2026 Vienna, Austria.
Prof. Wray Buntine's Visit
Burwood Corporate Center, Nov 2024
People
Academic Leads
- Son Tran Academic Lead
- Thommen George Karimpal Academic Lead
PhD Students
- Uzma Shafiq
- Mahla Tayarani Zare
- Arezoo Mousavi
- Elaheh Karimi Zarandi
- Dong Ming Zheng
- Salma Noor
Research Assistants
- Daniel EE Gan
- Shashwat Suthar
- Anjum Asiya
- Tracey Mai
- Anagha Varma
External Collaborators
- Joy Dhar
- Manish Pandey
Poster Wall
A selection of posters presented by the DMKD Group at conferences, workshops, research retreats, and project meetings.
Research streams
The DMKD Group works across five connected research streams. Each stream focuses on a different aspect of data mining, knowledge discovery, and machine learning, while sharing a common interest in structured data, knowledge, robustness, and practical impact.
Generative Tabular Models
We develop generative models for structured tabular data, with emphasis on mixed data types, feature dependencies, fairness, privacy, robustness, data scarcity, and reliable synthetic data generation.
Sample publications
- Ensembled Bayesian Tabular Data Generator, Knowledge and Information Systems, 2026.
- MEG: Masked Ensemble Tabular Data Generator, ICDM, 2023.
- GANBLR++: Incorporating Capacity to Generate Numeric Attributes and Leveraging Unrestricted Bayesian Networks, SDM, 2022.
Knowledge-guided Learning
We study how domain knowledge can be injected into machine learning through knowledge graphs, constraints, semantic relations, scientific principles, engineering models, architectures, losses, and inference procedures.
Sample publications
- KGGAT: Knowledge-Guided Graph Attention Network for Multi-Label Image Classification, CVPR Findings, 2026.
- Knowledge-Guided Graph Convolutional Network for Multi-Label Image Classification, ICRA, 2026.
- Knowledge-Guided Object Detection via Bayesian Networks and Knowledge Graphs, Expert Systems with Applications, 2026.
Multimodal Learning
We develop models that learn from multiple sources of information, especially in medical image analysis. This stream includes work on multimodal fusion, hybrid attention, geometry-aware learning, foundation-model adaptation, uncertainty, and robust medical imaging.
Sample publications
Anomaly Detection
We investigate anomaly detection for structured and tabular data, especially cases where anomalies are not merely isolated points but records that violate normal dependency structure. Current work studies struture-aware anomaly detection for heterogeneous tabular data.
Sample publications
- Coming Soon
Tabular Classification
We develop classification methods for tabular data, with emphasis on feature interactions, discretisation, Bayesian network classifiers, logistic regression, feature engineering, causal learning, robustness, and interpretability.
Sample publications
- Effective Interpretable Learning for Large-Scale Categorical Data, Data Mining and Knowledge Discovery, 2024.
- Improving Neural Network’s Robustness on Tabular Data with D-Layers, Data Mining and Knowledge Discovery, 2024.
- Scalable Learning of Bayesian Network Classifiers, Journal of Machine Learning Research, 2016.
Prospective Students and Visitors
Prospective PhD students, visiting researchers, and collaborators interested in generative tabular models, knowledge-guided learning, multimodal learning, anomaly detection, tabular classification, causal learning, or applied data science are welcome to contact me with a short research statement, CV, transcripts, and links to relevant publications or code.