Xiaoxiao Li

Faculty of Applied Science

Research Interests

Machine Learning

Deep Learning

Explainable AI

Trustworthy AI

Privacy and Security

Medical image analysis

Bioinformatics

Relevant Thesis-Based Degree Programs

View all programs

Graduate Student Supervision

Doctoral Student Supervision

Dissertations completed in 2010 or later are listed below. Please note that there is a 6-12 month delay to add the latest dissertations.

Enhancing privacy and efficiency in AI models through data-oriented learning strategies (2025)

The rapid development of Artificial Intelligence (AI) is transforming how we work, live, and innovate. Despite achieving impressive performance, AI models face challenges in reliability, privacy, and efficiency for real-world deployment. This thesis addresses these concerns through two main objectives: 1) tackling data privacy and heterogeneity via federated learning (FL), and 2) auditing data usage and quality during both pre-training and post-training stages. Efficiency is also emphasized to ensure practical applicability.All solutions in this thesis share a data-centric view, grounded in the belief that AI models are shaped by data and can, in turn, reveal information about it. To enhance privacy and efficiency, we first improve heterogeneous FL with new frameworks such as FedLGD and DeSA. These methods use virtual data to overcome challenges like heterogeneity, asynchronization, and decentralized coordination, leading to more robust and scalable FL systems. Both theoretical analysis and empirical results support their effectiveness.Then, we shift to data auditing, critical for ensuring data quality and ethical use. For post-training analysis, we introduce EMA, a method that combines membership inference with statistical ensembles to verify training data legitimacy. For pre-training assessment, EXAMINE leverages self-supervised learning to evaluate data quality. These methods enhance the robustness and accountability of AI systems, particularly in sensitive fields like healthcare.Finally, we propose future research directions that aim to improve scalability and adapt our work to emerging model architectures, enabling broader applicability in resource-limited and multimodal settings.In summary, this thesis offers a comprehensive approach to building trustworthy and efficient AI systems through innovations in federated learning and data auditing. By adopting a data-oriented perspective, it addresses key challenges in privacy, scalability, and accountability. The proposed methods, validated both theoretically and empirically, demonstrate strong potential for real-world impact, especially in ethically sensitive domains.

View record

Master's Student Supervision

Theses completed in 2010 or later are listed below. Please note that there is a 6-12 month delay to add the latest theses.

Overcoming catastrophic forgetting in heterogeneous federated learning (2025)

Federated learning (FL) has gained prominence in the medical domain for enabling collaborative machine learning across decentralized datasets while preserving patient privacy. However, key challenges such as catastrophic forgetting and local overfitting hinder the effectiveness of FL models in heterogeneous healthcare environments. This thesis addresses these challenges through two complementary studies. The first study introduces Federated Impressions, a synthetic data generation method that extracts the knowledge of the global model without compromising privacy. By training the local model on these impressions, the method mitigates catastrophic forgetting, allowing the model to retain critical knowledge while generalizing effectively across diverse datasets. This approach ensures robustness and adaptability to evolving medical data distributions. The second study enhances FL for medical imaging by leveraging Vision Transformers (ViTs) in a DualPrompt architecture. Prompts are divided into shared and group prompts: shared prompts capture generalized medical features and are trained exclusively on Federated Impressions alongside local data to prevent overfitting, while group prompts adapt to specific client tasks using local data. This method effectively mitigates catastrophic forgetting, preserving personalization through group prompts while maintaining global generalization via shared prompts. It strikes a balance between retaining global medical knowledge and adapting to localized data distributions. Experimental evaluations on benchmark medical datasets demonstrate that the proposed approaches improve global generalization by avoiding local overfitting. These advancements make FL more robust and effective in heterogeneous medical environments, addressing critical challenges in privacy, scalability, and catastrophic forgetting. This thesis lays the groundwork for scalable and adaptive FL frameworks tailored to medical applications, fostering innovation in privacy-preserving healthcare AI.

View record

Towards trustworthy and explainable AI: applications in healthcare (2025)

Differentially private neural tangent kernels for privacy-preserving data generation and distillation (2024)

Quantifying motor vigour in Parkinson’s disease using EEG: a dual workflow of orthogonal multiple instance learning and ICA-based analysis (2024)

Incremental learning and federated learning for heterogeneous medical image analysis (2023)

Standard deep learning paradigm may not be practical over real-world heterogeneous medical data, where new disease merges over time with data acquired in a distributed manner across various hospitals. There have been approaches to facilitate the training of deep models over two primary categories of heterogeneity, including 1) class incremental learning, which offers a promising solution for sequential heterogeneity by adapting a deep network trained on previous disease classes to handle newly introduced diseases over time; 2) federated learning, which offers a promising solution for distributed heterogeneity, by training a global model on a centralized server over private datasets of various hospitals or clients without requiring them to share data. The core challenge in both approaches is catastrophic forgetting, which refers to performance degradation on previously trained data when adapting a model to available data. Due to strict patient privacy regulations, storing and sharing medical data are often discouraged, posing a significant hurdle in addressing such a forgetting. We propose to leverage medical data synthesis to recover inaccessible medical data in heterogeneous learning, presenting two distinctive novel frameworks. Our first framework introduces a novel two-step, data-free class incremental learning pipeline. Initially, it synthesizes data by inverting trained model weights on previous classes and matching statistics saved in continual normalization layers to obtain continual class-specific samples. Subsequently, the model is updated by incorporating three novel loss functions to enhance the utility of synthesized data and mitigate forgetting. Extensive experiments demonstrate that the proposed framework achieves comparative results with state-of-the-art methods on four public MedMNIST datasets and an in-house heart echocardiography dataset. We propose our second framework as a novel federated learning approach to mitigate forgetting by generating and utilizing united global synthetic data among clients. First, we proposed constrained model inversion over the server model to enforce an information-preserving property in synthetic data and leverage the global distribution captured in the globally aggregated server model. Then, we utilize this synthetic data alongside the local data to enhance the generalization capabilities of local training. Extensive experiments show that the proposed method achieves state-of-the-art performance on the BloodMNIST and Retina datasets.

View record