Machine learning has the potential to change agricultural science to address pressing global challenges like food security, climate change, and sustainable farming. Agriculture sits at the intersection of complex systems—environmental, economic, and technological—drawing on diverse data from satellites, UAVs, ground sensors, and simulations. While these data sources offer immense potential for advancing scientific understanding and decision-support, they present key challenges for machine learning: synthesizing heterogenous data sources, working with limited labeled data, and ensuring results are accurate, trustworthy, explainable, and practical.
Our lab intends to establish machine learning as a key enabler for agricultural science advancement, creating both reliable, efficient models with broad applicability and deeper scientific insights. We aim to develop explainable, uncertainty-aware models that incorporate domain knowledge to deliver accurate, plausible, and reliable results. Further, we focus on the use foundation models and adaptation strategies to build systems that can work across diverse agricultural scenarios, creating scalable and robust solutions for the evolving challenges of digital agriculture.