Lianhua Chi

Research Dashboard

Research focused on Scalable AI and AI Autism.

My work sits at the intersection of machine learning, data mining, graph analytics, anomaly detection, and healthcare AI, with two main research themes: Scalable AI and AI Autism.

Google Scholar

Research Program 01

Scalable AI

This program develops scalable, adaptive, and interpretable AI for dynamic graph streams and evolving networked data. The emphasis is on compact representation design, real-time graph understanding, and robust decision support at scale.

Core Capabilities

  • Single-pass processing of dynamic graph streams
  • Compact graph representations for evolving networks
  • Adaptive learning under graph dynamics and drift
  • Reliable and interpretable scalable intelligence

Methodological Strengths

  • Adaptive hashing and sketching
  • Compact stream summarisation
  • Efficient graph representation learning at scale
  • Reliability and interpretability analysis

Selected Publications

Hashing for Adaptive Real-Time Graph Stream Classification With Concept Drifts

IEEE Transactions on Cybernetics, 2017.

Hashing Techniques: A Survey and Taxonomy

ACM Computing Surveys, 2017.

Fast Graph Stream Classification using Discriminative Clique Hashing

PAKDD, 2013. Best Paper Award.

Why It Matters

The agenda supports responsible, scalable, and interpretable AI for evolving networks across communications, finance, healthcare, transport, and public-sector data systems.

Research Program 02

AI Autism and Multimodal Neurodevelopmental Screening

This program focuses on earlier and more reliable autism screening by integrating behavioural and brain-based signals with modern multimodal representation learning and healthcare AI translation.

Current Directions

  • Behavioural screening AI
  • Multimodal neuroimaging for autism diagnosis
  • Brain and behaviour integration
  • Trustworthy and inclusive healthcare AI

Translation Focus

  • Early screening rather than retrospective analysis
  • Clinically useful multimodal fusion
  • Representation learning for real-world pathways
  • Cross-disciplinary health AI collaboration

Selected Work

DeepMNF: Deep Multimodal Neuroimaging Framework for Diagnosing Autism Spectrum Disorder

Artificial Intelligence in Medicine, 2023.

Autism Screening Using Deep Embedding Representation

International Conference on Computational Science, 2019.

Current multimodal AI agenda

Integrating brain and behavioural signals for translation-oriented screening.

Program Positioning

The program is designed to bridge algorithm development, autism-specific expertise, and real-world implementation rather than treating healthcare AI as a purely technical benchmark problem.

AI Autism Network

Community, autism research, multimodal AI, and national research partners.

The collaboration network is structured to support earlier autism screening through combined strengths in translation, diagnosis, multimodal modelling, and broader AI for medicine capability.

Community Translation

Little Supermen

Child and family-facing translation pathways and community engagement.

Visit collaborator

Autism Identification and Diagnosis

OTARC

Autism-specific research, diagnosis, and knowledge translation expertise.

Visit collaborator

Multimodal AI and Clinical Translation

AIM for Health Lab

Monash-based expertise in multimodal AI and clinically oriented health AI research.

Visit collaborator

Medical Innovation and Health Research AI

ACAMI

Bespoke AI for biomedical and health research with wider medical innovation reach.

Visit collaborator

National Research and Translational AI

CSIRO

National-scale research collaboration connecting scalable AI capability with translational impact.

Visit collaborator

Broader AI and Medicine Capability

THETA Team

Cross-institutional biomedical AI collaboration and translational ecosystem support.

Visit collaborator

Collaboration Logic

Together these collaborators support the full pathway from autism-specific expertise and community relevance to multimodal modelling, scalable AI capability, and broader medical AI translation.

Graph Anomaly Detection

Self-supervised and contrastive graph anomaly learning

Contributions in anomaly-aware graph learning connect self-supervised objectives, multi-scale contrastive methods, and dynamic graph analytics.

Medical AI

Healthcare machine learning beyond autism

Work extends to multimodal neuroimaging for Alzheimer's disease and other clinically grounded AI problems, reflecting broader medical AI capability.

Temporal and Spatial Data

Time-series and event-driven anomaly detection

Research on temporal modelling and multivariate anomaly detection connects scalable AI with wider dynamic data settings.

Language and Social Media

Geolocation prediction and social media mining

Earlier work in NLP and social media analytics adds depth in representation learning and data mining across heterogeneous information sources.

Overall Profile

Across these themes, the consistent focus is to design AI methods that remain useful when data are dynamic, noisy, large-scale, or high stakes.