|
|
James T. Lo
|
 |
Ph.D. 1969, University of Southern California, USA
Professor
Office: Room MP432
Phone: 410-455-2432
Email: jameslo@umbc.edu
|
 |
Dr. Lo was a postdoctoral research associate at the Stanford and Harvard
Universities from 1969 to 1972. He joined UMBC in 1972. In 1992, he solved
the notorious nonlinear filtering problem in its most general setting and
got a best paper award for it. Subsequently, he has developed effective,
systematic and general methodologies for adaptive, accommodative and robust
processing (e.g., system identification and control, and filtering). The
approach used was the neural computation approach. These methodologies were
applied to active noise and vibration control and successfully tested on
experimental data. To eliminate the local-minimum problem in estimating
nonlinear regression models and training neural networks, Dr. Lo discovered
the convexification method of global optimization for general data fitting.
In recent years, he developed a new paradigm of neural networks, called
temporal hierarchical probabilistic associative memory. It has most
properties desirable of the common cortical algorithm long hypothesized by
neuroscientists, and is expected to have applications to detection and
recognition of multiple and hierarchical causes in spatial and temporal data
and to construction of cognitive architectures.
|
|
|
|
|
|
