Professor Jie Lu

Innovation:
Fuzzy Machine-Learning Systems

Associate Dean (Research Excellence), UTS Faculty of Engineering and Information Technology; PhD (IS), Curtin University of Technology

Professor Jie Lu of the University of Technology Sydney.

Professor Jie Lu is a researcher in fuzzy machine learning and data engineering. She has successfully integrated fuzzy systems into machine learning and data engineering to solve machine learning under data uncertainty, insufficiency and ever-changing data-stream settings, reducing decision-making blindness.

The challenges entailed determining how to precisely measure unforeseeable changes in streaming data distribution, known as concept drift, which reduces the data-driven prediction accuracy over time, as well as finding methods to establish an efficient way of assisting machine learning in data-insufficient situations when high volumes of data are required to train the model.

Lu, along with the students under her supervision, addressed these challenges by developing fuzzy competence models that indirectly measure changes in data distribution via changes in competence and providing innovative methods for concept drift detection, understanding and reaction, and by proposing fuzzy transfer learning algorithms to effectively transfer knowledge learned from one or more source domains to a target domain under uncertainty.

This pioneering research launched a new methodology and provided a guideline for data engineers to respond to situations involving fast-changing, insufficiently labelled and uncertain data.

Lu’s breakthrough study produced novel fuzzy competence models and algorithms that accurately detect concept drift in data streams and react to it accordingly.

Another of Lu’s studies pioneered fuzzy transfer learning to effectively transfer knowledge from a source domain to a target domain with insufficient data — new markets, for instance — or uncertain data for learning, particularly when the source and target domains have different feature spaces for regression.

Lu’s innovation was to convert numeric data to granular values in the form of fuzzy sets and to then map feature spaces in source and target domains by building a fuzzy latent feature space, transferring the knowledge of multiple heterogeneous source domains to target outputs.

This pioneering research launched a new methodology and provided a guideline for data engineers to respond to situations involving fast-changing, insufficiently labelled and uncertain data.

These achievements are reflected in Lu’s inaugural directorship of the University of Technology Sydney’s Centre for Artificial Intelligence.

MORE IN Community

Julian O’Shea

Solartuk Expedition

Dominic De Gioia

Beer with an Engineer