Part of a nationwide $22.5-million initiative

Brown receives $2.3M from NSF to study human and machine learning

Brown University was the largest New England recipient of the National Science Foundation's $22.5-million Learning and Intelligent Systems (LIS) grant program. Brown's award will fund three projects over three years.

PROVIDENCE, R.I. -- Brown University has received three research grants worth a total of $2.3 million from the National Science Foundation (NSF) in support of research into how humans learn and create. The grants are part of a nationwide $22.5-million Learning and Intelligent Systems (LIS) project announced yesterday (Thursday, Oct. 2, 1997) by the NSF.

Twenty colleges and universities across the country received 28 grants in support of interdisciplinary research projects. Brown, with three grants, was the second-largest recipient after Carnegie-Mellon University, which had four. LIS is the first leg of a three-component NSF program in Knowledge and Distributed Intelligence.

Brown's three research projects will involve faculty from the Departments of Cognitive and Linguistic Sciences, Computer Science, Neuroscience and Physics as well as from the Division of Applied Mathematics. Faculty from Cognitive and Linguistic Sciences will serve as principal investigators for the grants.

Editors: Principal investigators and their colleagues from the three research projects are available for interview through the Brown News Bureau.

"The projects run the gamut from language to object recognition to vision and neuroscience," said Sheila Blumstein, chair of cognitive and linguistic sciences. "We need to know that information to understand the nature of the human mind."

By studying basic functions of the mind like navigation, scientists may one day create robots able to solve the problems necessary to free themselves from obstacles on other planets. Studying the neural processes may lead to a greater understanding of visual images, machine translation and computers that can interact with humans in English.

Brown's three research project include "Structured Statistical Learning," "Adaptive Cortical Computation in the Visual Domain," and "Learning Minimal Representations for Visual Navigation and Recognition."

Structured Statistical Learning

Mark Johnson, principal investigator
$775,420

This grant will fund research on the fundamental mechanisms for learning the compositional structures of the basic cognitive tasks of language, vision, planning and motor behavior. It is under the direction of Mark E. Johnson, Department of Cognitive and Linguistic Sciences, and includes David Mumford and Stuart A. Geman, Division of Applied Mathematics; John P. Donoghue, Department of Neuroscience; and Eugene Charniak, Department of Computer Science.

The project involves three interacting lines of research. The first refines and extends statistical learning models; the second applies those models to language, vision and planning; and the third develops and applies new experimental and analysis techniques for probing the neural mechanisms that learn and exploit compositional structure. It will investigate, for example, the relationship between words and phrases in language, the way the images are comprehended as configurations of objects, and the processes by which a robot can learn the structure of its environment.

Adaptive Cortical Computation in the Visual Domain

James Anderson and Michael J. Tarr, principal investigators
$772,920

This grant will fund research into the functional interactions among groups of neurons in the brain - research that may provide insights for the design of artificial recognition and decision-making systems. It is under the direction of James A. Anderson and Michael J. Tarr, Department of Cognitive and Linguistic Sciences, and includes Michael A. Paradiso, Department of Neuroscience; and Gerald S. Guralnik, Department of Physics.

The project will include study of a neural network model called a "network of networks" to mimic the activity of the neural networks in the brain. The model will be tested for a greater understanding of what happens in normal brain function and when people suffer from brain damage or other deficits.

Learning Minimal Representations for Visual Navigation and Recognition

William H. Warren Jr. and Michael J. Tarr, principal investigators
$824,996

This grant will fund research into the intelligence exhibited in interactions among sensory-motor activities and cognitive capacities like reasoning, planning and learning - how, for example, humans and robots are able to navigate through complex environments. It is under the direction of William H. Warren Jr. and Michael J. Tarr, Department of Cognitive and Linguistic Sciences, and includes Leslie P. Kaelbling, Department of Computer Science.

The research will investigate the structure of "cognitive maps," how knowledge of a new environment is learned, and how it is actually used to navigate during locomotion. It may one day lead to the construction of robots that can navigate as well as humans.