Link to Session Platform

Chairs(s):

Nattapon Jaroenchai, University of Illinois at Urbana-Champaign

; Shaowen Wang, University of Illinois at Urbana-Champaign

Description:

Recent advancements in high-performance computing and hardware have resulted in several state-of-the-art machine learning approaches (e.g., decision tree learning, reinforcement learning, inductive logic programming, Bayesian networks, and clustering) that can be applied to geospatial analyses. Deep learning has become the hottest trend in geospatial applications, especially in the case of supervised deep convolutional neural networks, which have attracted great interest in the computer vision and image processing communities.

These advancements are driven by the increasing need to process large amounts of data generated by the ever-increasing availability of sensors in remote sensing. This unprecedented amount of data generated every day requires the use of AI for exploration and knowledge extraction.

Many AI algorithms, however, are still in their immaturity in terms of scientific understanding. For example, CNNs are often constructed through trial and error. Basic questions such as "How many layers should be utilized in total?" remain. While researchers have access to a vast library of diverse AI algorithms, AI must be employed in conjunction with physical principles and scientific interpretation. Therefore, this session focuses on further understanding the method for applying AI and deep learning in geospatial research such as algorithm development, data training strategies, and implementations.

Non-Presenting Participants Agenda