Machine Learning

Design of intelligent algorithms.

Machine learning is the science of interpreting and acquiring useful knowledge from data. In Spatial Analytics group, we use machine learning to make systems that see, sense, and interact with their environments. More specifically, we develop mathematical theory, design and implement algorithms that can process spatiotemporal data to detect objects, track and recover their shape and spatial layout, recognize their classes, and understand their actions.

Towards this goal, we build models that describe statistical relationships and patterns between the observed data and the world state. Certain models are determined by a set of parameters. In learning, we determine parameters that reflect the relationship between the data and the world, and then infer about the world state using the model and new data. In some cases, we apply data driven machine learning techniques such as dictionary learning to improve data quality and manage the complex nature of visual data to make subsequent information extraction more effective.

Machine learning for spatiotemporal data is a challenge as real-world problems are often subjectively defined, hard to model, and computationally intractable. Nevertheless, our motivation is that there is a working proof: amazing human perceptual system!