Artificial Intelligence

AI papers from UCI

Machine Learning

Our research involves the development and analysis of algorithms that identify patterns in observed data in order to make predictions about unseen data. New learning algorithms often result from research into the effect of problem properties on the accuracy and run-time of existing algorithms.

We investigate learning from structured databases (for applications such as screening loan applicants), image data (for applications such as character recognition), and text collections (for applications such as locating relevant sites on the World Wide Web). UCI also maintains the international machine learning database repository, an archive of over 100 databases used specifically for evaluating machine learning algorithms.

Knowledge Discovery and Data Mining

Databases with millions of records and thousands of fields are now common in business, medicine, engineering, and the sciences. The problem of extracting useful information from such data sets is an important practical problem. Research on this topic focuses on key questions such as how can one build useful descriptive models that are both accurate and understandable? Probabilistic and statistical techniques in particular, play a key role in both analyzing the inference process from a theoretical viewpoint and providing a principled basis for algorithm development. Ongoing projects include the integration of image and text health-care data for finding diagnostic rules, automated analysis of time-series engineering data from the Space Shuttle, and discovery of recurrent spatial patterns in historical pressure records of the Earth's upper-atmosphere.

Automated Reasoning; Constraint networks and Probabilistic networks.

Automated reasoning investigates methods by which knowledge is represented and used to emulate human-like thought processes. Although most reasoning tasks were found to be computationally hard, it is believed that approximation methods based on tractable models can effectively cover a significant portion of intelligent activities. Accordingly, research at UCI is focused developing flexible and expressive representations that accommodate efficient reasoning, by:

• identifying structured knowledge-domains that are tractable for reasoning,
• compiling knowledge-bases into such tractable structures,
• developing reasoning algorithms that will be effective in practice,
• exploiting parallelism and distributive computation, inspired by brain architecture.

Research at UCI has focused on constraint networks and probabilistic networks as the primary models for addressing these issues. These frameworks unify and cut across many traditional areas in Artificial Intelligence. Constraint processing is a paradigm for formulating knowledge in terms of a set of existing or desired relationships among entities, without specifying methods for achieving such relationships. A variety of constraint processing techniques have been developed and applied to diverse tasks such as vision, design, diagnosis, truth maintenance, scheduling, default reasoning, spatio-temporal reasoning, logic programming, and user interface. Belief networks provide a formalism for reasoning about partial beliefs under conditions of uncertainty. They capture causal influences between linked variables and are widely applicable for knowledge diffusion, diagnosis, abduction and planning. The current focus at UCI is on application areas such as diagnosis, planning and scheduling and on probabilistic decoding.

Faculty working in Automated reasoning

Useful links

Brain Modeling

Biomedical Computing: Computational Biology and Medical Informatics

ICS faculty are involved in a project to develop a knowledge-based systems for recommending a customized multiple-drug therapy for HIV infected patients. We are also exploring an approach to creating knowledge-based systems by learning from patient data and have identified guidelines for screening for forms of dementia such as Alzheimer's disease. We have developed knowledge-based approaches to protein structure prediction, and implemented novel sequence-structure search algorithms and recognition methods. We are modeling DNA mutation and repair in connection with cancer-related studies.

The biomedical domain is a rich source of application problems on which to test AI methods, and we welcome inquiries from academic and industrial colleagues with interesting biomedical research problems amenable to an AI-based computational approach. AI methods are applicable to many of the questions in computational biology.

Faculty