Paul E. Utgoff - US grants

Affiliations:

University of Massachusetts, Amherst, Amherst, MA

Area:

Computer Science

Tree Info Publications Similar researchers PubMed Report error

We are testing a new system for linking grants to scientists.

The funding information displayed below comes from the NIH Research Portfolio Online Reporting Tools and the NSF Award Database.
The grant data on this page is limited to grants awarded in the United States and is thus partial. It can nonetheless be used to understand how funding patterns influence mentorship networks and vice-versa, which has deep implications on how research is done.
You can help! If you notice any innacuracies, please sign in and mark grants as correct or incorrect matches.

Sign in to see low-probability grants and correct any errors in linkage between grants and researchers.

High-probability grants

According to our matching algorithm, Paul E. Utgoff is the likely recipient of the following grants.

Filter high-probability grants:

Years	Recipients	Code	Title / Keywords	Matching score
1987 — 1989	Utgoff, Paul Barto, Andrew (co-PI) [⬀]	N/AActivity Code Description: No activity code was retrieved: click on the grant title for more information	Learning Efficient Recognizers For Analytically Derived Concepts (Computer and Information Science) @ University of Massachusetts Amherst This research develops artificial intelligence methods for the automatic formulation and use of concepts by computer systems. In particular, techniques which enable a machine to develop concepts based on its ability to "explain" instances presented to it are combined with methods to rapidly "classify" new instances into one of its learned categories. The goal is to convert the machine's inefficient but correct "explanation" procedures into efficient classification routines. Methods used will include both symbol-processing strategies and newer, "connectionist" approaches. The importance of this research is that exploratory artificial intelligence programs, now capable of some limited learning, must be substantially improved. In particular, artificial intelligence systems must be developed to both learn new categories (concepts) efficiently and apply these categories rapidly when presented with a high rate of data and new experience. Show summary Hide summary	0.915
1993 — 1996	Utgoff, Paul	N/AActivity Code Description: No activity code was retrieved: click on the grant title for more information	Dynamic Recursive Representation Selection For Inductive Learning @ University of Massachusetts Amherst The automation of classification learning is an ongoing concern in the machine learning, pattern recognition and statistics communities. People are very good at designing and collecting data, but less good at analyzing it and extracting useful generalizations. It is necessary to improve the techniques for automatic data analysis. The state of the art in inductive learning is that there exist several dozen inductive learning algorithms/concept representations, each of which is best on some proper subset of the possible inductive learning tasks. The project will develop an efficient method for selecting the best algorithm/concept representation for each recursive call of a divide-and-conquer tree induction algorithm. The combination of different representations into a hybrid significantly improves the ability of the inductive learning program to find a highly accurate generalization of the data. The major objective of this research will be an approach to learning that includes an efficient search for an appropriate hybrid representation, guided by detection of specific forms of hypothesis pathology. The creation of systems that search multiple representation spaces greatly increases the autonomy of learning machine, which will make data analysis techniques accessible to a wider range of scientists.// Show summary Hide summary	0.915
1997 — 2001	Utgoff, Paul	N/AActivity Code Description: No activity code was retrieved: click on the grant title for more information	Constructive Function Approximation @ University of Massachusetts Amherst The goal of this research project is to create new algorithms for numerical function approximation that are particularly suited to value function approximation in reinforcement learning. These algorithms localize the approximation error in the domain of the function, and respond by constructing features that enable further error reduction. It is essential that these algorithms be able to approximate well even when the function appears to be changing as learning progresses. The results of this research will be algorithms that reduce error in a repeatable manner, and that produce an approximation that is inspectable and understandable. Application tasks for the research will include instruction scheduling and autonomous agent policy learning. The research aims to enable practioners who employ automatic learning methods to achieve more accurate and more understandable results with less human engineering. Show summary Hide summary	0.915
2001 — 2004	Raphael, Christopher [⬀] Utgoff, Paul Sebastiani, Paola (co-PI) [⬀]	N/AActivity Code Description: No activity code was retrieved: click on the grant title for more information	Itr/Im (Cse): Musical Signal Recognition @ University of Massachusetts Amherst The World Wide Web contains a vast and ever-growing collection of music audio files representing nearly every musical style, ensemble, genre, country, culture, and time period. However, with the exception of the information conveyed in the title, the contents of such audio files can only be understood by listening to the files. Thus searches of audio files analogous to those performed by text-based search engines are currently impossible. In this project the PI will study and implement solutions to the "Signal to Score" problem in which an audio file is transcribed into a format capturing information similar to that contained in a printed musical score. The PI's approach splits the task into two components: "Signal to Piano Roll" in which the musical signal is transcribed into a MIDI-like representation, and "Rhythmic Parsing" in which the piano roll representation is further transcribed into a musical score or equivalent representation. The goal is to allow the generation of searchable data bases that contain high level music descriptions, which could be used to algorithmically answer questions on musical content such as "Is the audio file likely to be a blues song?" or "What is the time signature of the music?" Show summary Hide summary	0.915
2001 — 2005	Utgoff, Paul	N/AActivity Code Description: No activity code was retrieved: click on the grant title for more information	Feature Construction For Large Discrete Domains @ University of Massachusetts Amherst IIS-0097218 Paul E. Utgoff University of Massachusetts at Amherst $119,407 - 12 mos Feature Construction for Large Discrete Domains This is the first year funding of a three year continuing award. This project focuses on the problem of learning the representations on which more widely studied data fitting programs depend. Thus, instead of providing a representation and watching the fitting algorithm run its course, the goal is to attack the more fundamental problem of learning the representation itself. The objective is to produce an agent that can identify all of the useful features of a large discrete domain. Artificial neural networks of the typical one or two layers of hidden units cannot scale to large problems of the kind that the PI wants to solve (learning in large discrete domains), because this kind of `few-layered learning' suffers not only from local minima and shallow gradients, but more fundamentally from inappropriate bases and the inherent need for exponentially many features (hidden units) imposed by the constraint of so few layers of features. The PI will pursue scalable methods that he characterizes as `many-layered learning', which will move the state-of-the-art past the current nonscalable practices of feature construction, which in turn will affect much of the work in function approximation, including the current nonscalable use of few-layered artificial neural networks. The project will produce a variety of algorithms for many-layered learning. Among them will be one for building a nested feature representation based on problem-solving experience. A second will demonstrate that knowledge layering and decomposition follow naturally from using only simple learning mechanisms to learn the next most easily learned features based on the representation learned thus far. The larger implications for many-layered learning on intelligence will be investigated. Show summary Hide summary	0.915
2003 — 2010	Utgoff, Paul Schultz, Howard Hanson, Allen [⬀] Riseman, Edward (co-PI) [⬀]	N/AActivity Code Description: No activity code was retrieved: click on the grant title for more information	Collaborative Research: Itr: Interactive Software Systems For Expert-Assisted Image Analysis and Classification of Aquatic Particles @ University of Massachusetts Amherst This project involves research on methodologies and algorithms for the automated and semi-automated classification and counting of particles in the upper ocean. The work will harness ideas from computer vision, machine learning, and image analysis to extract features from several different types of optical instruments used to study planktonic marine organisms and detritus, and then classify them so that they can be automatically recognized, sorted and counted. The goal is to develop systems that will automatically classify the majority of marine particles and flag the unusual or difficult-to-recognize features for human interpretation. Feature classification systems will be developed around three different classes of instruments that yield images of three different types of particles: epifluorescence microscopy (bacteria and nanoplankton), FlowCam (phytoplankton including types responsible for harmful algal blooms), in situ cameras (zooplankton and detritus). A variety of image classification methods will be investigated including ones based on neural networks and support vector machines. Reference images will be made from phytoplankton from the collection of the Center for Culture of Marine Phytoplankton and from plankton in samples taken from the Gulf of Maine. Images recorded in past expeditions and on at-sea expeditions funded elsewhere will also be used. As part of the project, a pattern recognition package will be added to the Phytopia educational multi-media CD-ROM used in schools and colleges. Show summary Hide summary	0.915