Meet Hassan Saadatmand: Developing Efficient AI for Time-Series Data
Hassan Saadatmand is a PhD student at Monash University and a member of the Temporal Analytics Lab.
His research focuses on developing efficient deep learning models for time-series classification, with a particular interest in how AI can capture complex temporal patterns while remaining efficient.
Hassan brings a broad research background spanning soft computing, machine learning, AI, feature selection, robotics and control engineering. His experience led to a growing interest in how intelligent systems can learn meaningful patterns from complex and sequential data, ultimately drawing him towards time-series analysis and sequence modelling.
His current PhD research, supervised by Dr Mahsa Salehi, Prof Hamid Rezatofighi and Prof Geoff Webb investigates modern sequence-learning approaches including structured state space models. A key focus is understanding how these models can capture long-range temporal dependencies while improving efficiency and scalability.
This work contributes to the broader goals of the Temporal Analytics Lab by exploring practical approaches to analysing temporal data, particularly in environments where computational and memory resources may be limited. Hassan is particularly interested in applications across healthcare and energy, where efficient analysis of time-series data could support improved monitoring, prediction and decision making.
One of Hassan’s research achievements to date is his publication in IEEE Transactions on Pattern Analysis and Machine Intelligence (TPAMI), where he developed a many-objective evolutionary approach to feature selection for high-dimensional and imbalanced classification problems. The work brought together several of his research interests, including evolutionary optimisation, machine learning and feature selection.
The research also reinforced an important principle that continues to influence Hassan’s work: machine learning models need to balance more than accuracy alone. Factors such as efficiency, complexity, interpretability and robustness can be equally important when developing models for real-world applications.
More recently, his research on MS4N, a structured state space model for time-series classification, has explored how relatively simple architectural improvements can improve the representation of complex temporal patterns without relying on increasingly complex models.
Through his PhD research, Hassan hopes to contribute to the development of AI systems that are accurate, efficient and scalable, helping to make advanced AI more practical for real-world applications and resource-constrained environments.