Lecture: 'AI Sampling of Molecular Paths Using Human Generated Trajectories in iMD-XR: working at the intersection of human-computer interaction, machine learning, and computational chemistry'
Molecular dynamics simulations often struggle to efficiently explore rare, high-value transition pathways. The rugged energy landscapes and long timescales involved mean that naive sampling wastes enormous computational resources on uninteresting regions of configuration space.
Interactive Molecular Dynamics in Virtual Reality (iMD-VR) offers a compelling alternative: human intuition, guided by real-time force feedback and immersive 3D visualization, can rapidly identify plausible reaction coordinates and transition paths that traditional algorithms miss.
This talk presents a framework for distilling that human expertise into AI systems capable of autonomous path sampling, exploring how trajectories captured from human operators navigating molecular systems in VR are used to train models that can later interact with iMD-VR simulations, and considering the practical challenges of this approach: extracting meaningful low-dimensional representations from high-dimensional VR interaction data, handling the sparsity and variability of human demonstrations, and validating that AI-generated paths respect the underlying physics rather than merely mimicking human motion patterns.
About the speaker
Mohamed Dhouioui is a postdoctoral researcher at CiTIUS (Research Centre on Intelligent Technologies, University of Santiago de Compostela). His current research focuses on the development of artificial intelligence systems for molecular dynamics simulations.
His research interests include machine learning, data science, molecular dynamics, and embedded systems. He holds a PhD in Computer Systems Engineering from the University of Sfax, where his doctoral research focused on Edge AI. His current work explores the application of AI to interactive molecular simulations, including human-generated trajectories and imitation learning in high-dimensional molecular systems.
Molecular dynamics simulations often struggle to efficiently explore rare, high-value transition pathways. The rugged energy landscapes and long timescales involved mean that naive sampling wastes enormous computational resources on uninteresting regions of configuration space.
Interactive Molecular Dynamics in Virtual Reality (iMD-VR) offers a compelling alternative: human intuition, guided by real-time force feedback and immersive 3D visualization, can rapidly identify plausible reaction coordinates and transition paths that traditional algorithms miss.
This talk presents a framework for distilling that human expertise into AI systems capable of autonomous path sampling, exploring how trajectories captured from human operators navigating molecular systems in VR are used to train models that can later interact with iMD-VR simulations, and considering the practical challenges of this approach: extracting meaningful low-dimensional representations from high-dimensional VR interaction data, handling the sparsity and variability of human demonstrations, and validating that AI-generated paths respect the underlying physics rather than merely mimicking human motion patterns.
About the speaker
Mohamed Dhouioui is a postdoctoral researcher at CiTIUS (Research Centre on Intelligent Technologies, University of Santiago de Compostela). His current research focuses on the development of artificial intelligence systems for molecular dynamics simulations.
His research interests include machine learning, data science, molecular dynamics, and embedded systems. He holds a PhD in Computer Systems Engineering from the University of Sfax, where his doctoral research focused on Edge AI. His current work explores the application of AI to interactive molecular simulations, including human-generated trajectories and imitation learning in high-dimensional molecular systems.
On-site event
Thursday, October 8, 2026
1791417600000
/events/lecture-ai-sampling-of-molecular-paths-using-human-generated-trajectories-in-imd-xr-working-at-the-intersection-of-human-computer-interaction-machine-learning-and-computational-chemistry
events_en