Professor of Machine Learning & Signal Processing
Professor Kreutz-Delgado's current research is concerned with the development of biologically inspired sensorimotor intelligent learning systems that can effectively function in unstructured, nonstationary environments and provide insight into human information processing and neuropathological disorders, such as Parkinson's disease. He is the lead researcher responsible for the development of EEG-based neurocomputation (NCP) signal processing algorithms for the US Army Research Laboratory's Cognitive and Neuroergonomics Collaborative Technology Alliance [CAN-CTA], an Army funded consortium of universities and research institutes. In furtherance of this activity, Prof. Kreutz-Delgado is affiliated with, and collaboratively interacts closely with researchers at, the UCSD Swartz Center for Computational Neuroscience [SCCN], which is responsible for managing the UCSD component of the Army CAN CTA research activity. Prof. Kreutz-Delgado is also a Co-PI on the recently funded collaborative NSF EFRI research activity, "Distributed Brain Dynamics in Human Motor Control." This activity involves the development of novel imaging methods to monitor and record body and brain activity during real-world tasks. The resulting data will be used to develop detailed, large-scale models of activity in the brain’s basal ganglia-cortical networks, where Parkinson’s disease (PD) takes its toll, with the ultimate goal of ameliorating the symptoms of PD.
These research activities involve the sophisticated use of statistical signal processing; statistical learning theory and pattern recognition; adaptive sensory-motor control; nonlinear dynamics and multibody systems theory; and optimization theory. Prof. Kreutz-Delgado's has experience in all of these areas. Before joining the faculty at UC San Diego, he was a researcher at the NASA Jet Propulsion Laboratory (JPL), California Institute of Technology (Caltech), involved with the development of adaptive, intelligent telerobotic systems for use in space exploration and satellite servicing and repair. His technical contributions in robotics include the development of a spatial operator algebra for the analysis and control of complex, multibody systems which exploits an algorithmic analogy to the recursions of discrete-time Kalman filtering and smoothing (work which resulted in a NASA Technology Achievement Award); the application of nonlinear dynamical reduction for robust sensory-motor control of multilimbed robotic systems; and the use of learning theory and differential topology for the development of trainable nonlinear representations for sensory-motor control. Beginning in the mid-1990's he began research into the sparse solution recovery problem and dictionary learning. Currently, sparse recovery and dictionary learning techniques are being further developed and applied to inferring cognitive function from various brain imaging modalities in support of the Army CAN CTA and NSF EFRI activities mentioned above. Prof. Kreutz-Delgado is also currently involved with research into the development of deep learning architectures combined with reinforcement learning as providing reasonable models of brain cognitive function for tasks such as situation awareness, object recognition and sequential gaming.
Distinctions and Current Professional Activities
update: September 18, 2013