Research Here and There
Zhizheng Zhang(张志政)
Southeast University(东南大学)
Here
There
Research
Mathematical
World
Problem
Models
Ontology
Languages
Interpretations
Algorithms
Outline
• Did
– Fuzzy relational database
– Financial Time Series Forecasting
– Preference Logic
• Doing
– Learning ASP/Action Languages
– Medical decsion support system
– Modeling interaction scenario.
• To do
– Formalizing medical guidelines using action language
– Logic for interaction
• Logic programs sequences
Did
• Fuzzy Relational Database
– World: linguistic variable+fuzzy raletions
– Language: fuzzy SQL
– Model: fuzzy relations theory
– Problem: Fuzzy Query
• Financial Time Series Forecasting
– World: exchange rate change history, concerning factors
(e.g. oil price.)
– Language: neural network, Baysian network, fuzzy
rules….
– Model: Time Series xt+1=f(xt, xt-1,…,yt,yt-1,…)
– Problem: What are the parameters’ values in a given
language for a given world of histories.
• My work on preference logic
– Provide an axiomization system with four
kinds of preferences introduced in
[1]. J. van Benthem, S. van Otterloo, and O. Roy. Preference logic,
conditionals, and solution concepts in games. In H. Lagerlund,S.
Lindström, and R. Sliwinski, editors, Modality Matters, 61–76, 2006.
[2]. Johan van Benthem. Games in dynamic-epistemic logic. Bulletin of
Economic Research, 53:219–248, 2001.
• ,  are propositional logic formula
– > : there exist at least one model m of  and at least
one model m’ of  such that m is better than m’.
– > : there exist at least one model m of  such that m is
better than every model of .
– > : there exist at least one model m of  such that every
model of  is better than it.
– > : for any model m of  and any model m’ of , m is
better than m’ .
• Satisfaction of a preference logic
formula
• Preference Revision
– Example, In valentine's day, based on your and your
girlfriend’s usual preferences, your new preferences are
romantic movie is better than other types of movies, and
more stars is better.
Doing
• Learning ASP/Action Languages
– Why:
• Motivation Problems:
– How to make new questions automatically according to a
student’s QA histories and textbook knowledge in
educational assistant systems?
– How to diagnose a student’s psychological status according
to his behaviors records and middle school students
psychological knowledge?
• We hope our programming languages are our natural
languages.
Medical decision support system
• World: dynamic domain
• Model: transition diagram
• Language: action language
• Problem: Planning
• Difficulties: complex ontology, conflict
knowledge, preferences.
• Modeling interaction scenario
– Static Games. The status of agents are static
– Dynamic Games. The behaviors of agents can
change their preferences, beliefs.
To Do
1. Formalizing some diseases prevention and
cure guidelines using action language
•
I don’t know what difficulties we will meet. Maybe
they include: uncertain, complex ontology, quantitative
constriants, commonsense……etc..
•
My question: Is “Modular” method good enough to
model the complex domain for action languages?
My thinking now: Yes, it is successful in C++ and jave, but need
to test it in the real applications by using it to model domains with
character of inheritance hierarchy.
• Logic for interaction
– Axiomatization
• If preferences are the basis of social choice, how to model the
decision procedures when preferences are dynamic.
– Computation issue of some interaction scenarios
• Formalizing interaction procedures using ASP/AL
– Why we have ASP sequences?
» Beliefs are changing.
» Prioritized goals
» We formalize the open/evolving domain step by step
» Every agent has its own belief on world, but they have to
work together.
• Question: What is the rational “answer set” of
ASP sequences?
My thinking:
1). Rational answer sets of the sequences should
also be those literals that we are forced to believe.
2). If the union of those ASP is consistent. GOOD!
3). Restoring consistency is rational if we can remove
inconsistency by its causes and making a minimal
change.
What are the causes of inconsistencies in
an ASP seqeuences? Example, P<Q
(1). Indirect Exceptions
P: {p:-c, not –p.
q:-p.}
Q: {c. –q}
CR-Prolog is great!
(2) not-cycle
P: {a:-b.}
Q: {b:-not a.}
(3)Unsatisfied Constraints
P: {:-a, b. a.}
Q: {b.}
(4) others?
P: {a.}
Q: {-a.}
(5) Any others?
How to deal with them?
THANK YOU