The Neural Networks and Machine Learning lab specializes in:
- Predictive Modeling
- Automated Learning
- Neural Networks
- Data Analysis for Knowledge Discovery
- Pattern/Audio/Facial Recognition
- Path Planning
The Perception, Control and Cognition laboratory blends deep neural networks with Bayesian models to create flexible, scalable inference algorithms that can be trained on input from the natural world. Specific focus areas include probabilistic programming, natural language processing, integrated cognitive systems, dynamical systems modeling, robotics, automated decision-making and optimal control.
The Applied Machine Learning Laboratory focuses on increasing the practicality of theoretical machine learning algorithms, especially as applied to autonomous robotic systems. We concentrate on making reinforcement learning techniques applicable to large problems, and on making it fast enough for on-line use. The lab is currently pursuing research projects in areas such as knowledge transfer, skill selection, skill composition, function approximation, multi-agent decomposition, and adversarial multi-agent learning. Q-learning, suitable function approximation such as RBF and back-propagation networks, and induced abstract models are the primary methodologies we use to solve problems.