mission of the machine learning and inference laboratory
The Machine Learning and Inference (MLI) Laboratory conducts fundamental and experimental research on the development of intelligent systems capable of advanced forms of learning, inference, and knowledge generation, and applies them to real-world problems. The mission of the laboratory is to contribute to the highest quality research and education in machine learning.
The focus of the laboratory is on hard machine learning problems faced in real world applications. These include dealing with complex data, often found in domains such as healthcare, extensively using existing knowledge in addition to data, combining multiple forms of learning, and producing results that are highly transparent and easily understandable by end users, not necessarily trained in machine learning.
what is machine learning?
A computer system learns if it improves its performance or knowledge due to experience, or if it adapts to a changing environment. The experience can be of the system that learns, or can be provided from outside, for example, in the form of data from which the system learns. Although the majority of machine learning methods concern learning from data, this is not the only available form of learning. Results of machine learning are in the form of knowledge or models (functions) representing what has been learned, and are most often used to make predictions about the future or unknown cases/situations or to help perform the system’s tasks better.
Major current and past research areas in MLI include: theory and computational models of learning and inference, data mining and knowledge discovery, machine learning and natural induction, inductive databases and knowledge scouts, behavior modeling and computer intrusion detection, non-Darwinian evolutionary computation, multistrategy learning and knowledge mining, intelligent systems for education, models of human plausible reasoning, and machine vision with learning capabilities.The developed methods and systems are experimentally applied to a wide spectrum of problems in science and industry.
Application areas of special interest include complex engineering design and optimization, computer security, medicine, health informatics, bioinformatics, earth sciences, sociology, biochemistry, communication networks, geographic information systems, world economy, computer vision, education and software engineering. The Laboratory supports education, scholarship, and research in these areas. It has a highly international team of researchers and the state-of-the-art computer facilities.
a brief history for the curious
The Machine Learning and Inference Laboratory (MLI) is one of the most active and accomplished laboratories in the world in the fields of machine learning and computational intelligence. It traces its roots to the Intelligent Systems Group (ISG) that Dr. Michalski established at the University of Illinois at Champaign-Urbana in 1972.
Since then, Dr. Michalski and his collaborators have been continuously publishing original research results in such areas as inductive learning, pattern recognition, advisory systems, machine inference, data mining and knowledge discovery, computer vision, plausible reasoning, and others. In 1980, they co-organized the first workshop in machine learning, and then several subsequent workshops/conferences in this field. These conferences now continue as International Conferences on Machine Learning, the most important conferences in this field.
In 1988, Dr. Michalski accepted an invitation from George Mason University (GMU) to move his research group there, and renamed it to the Machine Learning and Inference Laboratory (MLI). In 1991, MLI has co-organized the First International Workshop on Mulistrategy (MSL’91) in Harpers Ferry; and then subsequent MSL workshops (MSL’93 and MSL’96 in Harper’s Ferry, MSL’98 in Desenzano del Garda, Italy; and MSL’00 in Guimarães, Portugal). In 1992, it organized the first workshop on learning and vision. In 2008, MLI co-organized the first workshop on Computational Intelligence and Machine Learning Virtual Organizations.
In 2007, Dr. Janusz Wojtusiak became the director of MLI. His goal is to continue good traditions of high quality research, and explore new research areas and direction relevant to MLI. The special focus is on hard machine learning problems, including those found when analyzing hard problems found in healthcare.
MLI has pioneered several research directions, such as symbolic rule learning, variable-valued logic, constructive induction, conceptual clustering, variable-precision logic, multistrategy learning, learnable evolution model, inductive databases and knowledge scouts, learning from aggregated data, and others. With over 700 papers and 15 books and conference proceedings, MLI is one of the most published and continuously contributing new ideas research groups in the areas of its expertise.
summary of research and contributions
MLI has originated many innovative ideas and methods in the areas of machine learning, knowledge discovery, plausible inference, advisory systems, and computer vision. Among specific methodological contributions are: · Aq algorithm for solving complex covering problems that pioneered the progressive covering approach to inductive learning (a.k.a. separate and conquer), · variable-valued logic and attributional calculus· constructive induction · conceptual clustering· variable-precision logic (with Patrick Winston from MIT)· the core theory of human plausible reasoning (with Alan Collins from BBN)· multistrategy learning· inductive databases and knowledge scouts· learnable evolution model (LEM)· rule learning from aggregated data· computational intelligence and machine learning virtual organization (with Jacek Zurada from U. of Louisville).
To support education, MLI has established several graduate level courses and is providing computational and software resources to students seeking knowledge related to intelligent systems. In doing that MLI closely collaborates with the Health Informatics Learning Laboratory (HILL), Center for Discovery Science and Health Informatics, and various academic departments within and outside of GMU.
MLI organized numerous conferences and workshops, and has trained a large number of students and researchers in the areas of its activities. It has also been a place of frequent visits by international scholars and students from around the world.
main systems developed
Over the years, MLI has developed many experimental systems in the areas of machine learning, inference, knowledge discovery, and others:
· AQ family of rule learning systems (AQVAL, AQ11, AQ15, AQ18, AQ19, and most recently, AQ21 - current version under development)
· PLANT/s--the first agricultural expert system, and the first practical expert system with learning capabilities,
· INDUCE family for structural learning (INDUCE 1,..,4)
· CLUSTER family for conceptual clustering (CLUSTER 1, 2)
· ABACUS systems for equation discovery (ABACUS 1, 2)
· SPARC qualitative prediction systems (SPARC/E, SPARC/G)
· QUIN and ADVISE general purpose advisory systems
· MIST system for applying machine learning to vision
· INLEN family for multistrategy data mining and inference
· LEM systems for evolutionary computation, based on Learnable Evolution Model--a form non-Darwinian evolutionary computation, specially oriented for complex function optimization and engineering design.
On the invitation of the Boston Science Museum, our group has also developed ILLIAN system for demonstrating learning and discovery capabilities of computers for the public. The system toured major U.S. science museums. An extended version of this system, EMERALD-SUN, has been used in teaching AI/CI courses. A new implementation, iAQ, has been presented at many conferences and meetings, and is being distributed world-wide to academic and industrial organizations. It is available through internet (www.mli.gmu.edu/software.html).