Master Theme Machine Intelligence
Machine Intelligence is one of the themes
(specialisations) in the Computer Science Master Programme at the
University of Nijmegen. It relates closely to the research carried
out in the sections on
Intelligent Systems
and Model-based System Development.
Research
Machine Intelligence focuses on the development of systems that are
capable of handling complex problems due to their ability to take into
account formalised knowledge of a problem domain including its
underlying decision-making process. This knowledge is either
explicitly specified, and may then be acquired from humans
(human experts or specialised literature) or learnt from data, or it
is kept implicit, and then it is normally obtained from searching a
state space of possibilities. Some AI systems are also able to adapt
to their environment, i.e. they are adaptive.
Whereas Machine Intelligence focuses on the computer-based
aspects of intelligence, i.e., making computers behave
intelligently, the Artificial Intelligence master (for which the Social
Science Faculty is responsible, although we are involved in
this theme as well) has more
focus on the cognitive science aspects of intelligence,
i.e., trying to understand why humans behave intelligently
using paradigms from computing science. Below we make a
distinction between Machine Intelligence and Artificial
Intelligence, as these are different master programmes.
However, courses that are being offered as part of the
BSc programme, the term 'Artificial Intelligence' is used,
as the course are offered as part of the joint (Computing and Information
Sciences and Social Sciences) AI programme. Computing and information
bachelor students can take such courses as part of an AI minor
programme.
Fundamental research in Machine Intelligence deals with topics such
as:
- Knowledge representation: the development of special-purpose
languages tailored to the representation of various types of knowledge;
bridging the gap between cognition and formal systems of knowledge
representation
- Automated reasoning: methods that allow
reasoning with represented knowledge
- Machine learning: automatic discovery of knowledge from data
- Search: the design of fast, often heuristic,
algorithms to explore possibly large state spaces for a solution to a problem
In turn, these topics recur in various subfields of Machine Intelligence,
such as:
- Knowledge-based Systems: systems that consist of a knowledge-base,
containing formalised domain knowledge,
and an inference engine, consisting of reasoning algorithms,
that solve problems by reasoning with the
represented knowledge
- Intelligent Agents: systems that are constructed using the
metaphor of autonomous, distributed and communicating entities that
are able to solve problems
- Reasoning under Uncertainty: methods for representing and reasoning
with the uncertainty involved in decision making
Actual applications of Machine Intelligence research are found in various fields;
examples include: medical diagnostic expert systems, model-based
trouble shooting of cars and printing devices, decision-support
systems in management and industry, theorem proving systems in computer
science and mathematics, information retrieval on the Word Wide Web,
and data-mining in consumer markets.
In Nijmegen, research is focused on the following subthemes:
- Statistical Machine Learning and Data Mining (contact Tom Heskes, Elena Marchiori
or Peter Lucas):
statistical machine learning, optimization, probabilistic graphical networks and neural networks from data, extraction
of patterns from data, data-mining
- Bayesian Networks
(contact Peter Lucas,
Tom Heskes,
Bert Kappen or
Wim Wiegerinck):
graph-based representations of
probabilistic information, that support an attractive way of
reasoning with uncertain information in the light of available
evidence; forcus of research is on approximate inference and
knowledge representation
- AI and Bioinformatics
(contact Elena Marchiori):
heuristic optimization, efficient search and machine learning methods
for solving Bioinformaticsproblems;
actual applications include network analysis, metagenomics and gene
regulatory networks
- Model-based Reasoning Systems
(contact Peter Lucas):
these are systems that
use explicit models, e.g. cause-effect models, to solve problems
- Knowledge Representation (contact
Janos Sarbo):
cognitively based models of knowledge in various domains such as
natural language, logic, reasoning; practical applications of such models
in parsing, text summation, dialogue processing, and also in problem
specification and ontology definition
Courses
As a preparation to the Master's Theme, a number of AI courses are
also offered in as part of the BSc degree:
- Knowledge Representation and Reasoning: introductory AI course that introduces topics
such as state-space search, knowledge representation and automated reasoning,
knowledge-based systems, model-based reasoning, uncertainty, machine learning,
and knowledge engineering
- Learning and Reasoning Systems: course about
machine learning and data mining
It is also worth considering to take the minor artificial intelligence,
by which you supplement the above programme by a number of
other AI courses.
The following courses are offered as part of the Master's Programme in
Machine Intelligence:
Supplementary courses can be taken from the Master programme of AI as taught
as part of the AI Master programme at Donders Institute.
Examples are: