Invitation to RIAIoT Lunch Seminar Series on 25 March: “The Myopic Eye as an Accessible Model of Retinal Circuit Function and Neural Computation” & “Towards Interpretable AI for Molecular and Materials Science”

We are excited to announce the RIAIoT Lunch Seminar Series, featuring two engaging talks from our members: “The Myopic Eye as an Accessible Model of Retinal Circuit Function and Neural Computation” by Prof. Feng Pan (1st speaker) and “Towards Interpretable AI for Molecular and Materials Science” by Prof. Wanyu Lin (2nd speaker). This event will be held on 25 March, 2026, at The Hong Kong Polytechnic University.

Abstracts

• The Myopic Eye as an Accessible Model of Retinal Circuit Function and Neural Computation

Abstract: Myopia is thought to arise, at least in part, from altered retinal signaling, yet the biophysical mechanisms underlying this process remain poorly understood. Here, we investigate how the myopic mouse retina encodes focused and defocused images at the level of identified retinal circuits. We find that the myopic retina encodes visual information differently from the normal retina, and that lateral inhibitory networks play a critical role in detecting defocus and potentially driving myopia development. Using myopia as an example, we will discuss how retinal circuits—an integral part of the central nervous system—provide a uniquely accessible window into fundamental principles of neural computation.

• Towards Interpretable AI for Molecular and Materials Science

Abstract: 

Molecular and materials science are central to addressing global challenges in healthcare, energy sustainability, environmental protection, and next-generation technologies. Applications such as drug discovery, energy storage, carbon capture, catalyst design, and semiconductor development highlight the transformative potential of these fields. At the core of these advances is the ability to design and analyze complex molecular and material systems. AI for scientific discovery has therefore attracted growing interest across machine learning, physics, chemistry, and materials science. A key challenge is developing effective, efficient models of molecules and materials. Although deep learning can capture complex chemical and physical behavior, its “black‑box” nature often limits its ability to yield actionable scientific insights. This presentation underscores the essential role of interpretability in deep learning. By enhancing trust in model predictions and enabling the extraction of meaningful mechanistic understanding, interpretable AI frameworks empower scientists to uncover new principles and accelerate systematic discovery.