Abschlussarbeiten / Adviced Theses

Mögliche Themen

Theses topics are in the context of current research in the cognitive systems group.

Most topics can be adapted to bachelor or master/diploma level. Some topics are not suitable for computer science students but address psychology students (who select cognitive systems as elective). The pdf-files for the references to our own work can be found via www.cogsys.wiai.uni-bamberg.de/schmid/publications.html

We investigate example-driven algorithms for inductive synthesis of recursive functional programs from small sets of positive examples. Research topics include theoretical characterization of the approach, algorithmic realization, evaluation by example problems and comparison with alternative approaches as well as applications (for enduser programming support and automatization in the domain of automated theorem proving).

(1) Inductive Synthesis of Recursive Programs

Our approach to inductive program synthesis should be adapted for programming by demonstration. That is, user behavior in a text editor will be observed and query/replace commands which the user realized by hand (e.g. instead of using regex replace) will be inferred and proposed to the user.

  • Kitzelmann, E.; Schmid, U.: Inductive Synthesis of Functional Programs - an Explanation Based Generalization Approach. Journal of Machine Learning Research, 7(Feb), 429 - 454.


(1.1) Learning query/replace commands from enduser observations (DA/MA Level
only)

 

 

(2) Analogical Problem Solving and Learning

Programming by analogy can be seen as a special case of problem solving by analogy: A new programming problem is given either by examples or by specification and the (typically recursive) solution is generated by mapping the new problem to an already solved problem and transferring the solution.
In the context of the system IPAL we learn recursive program schemes. We propose an approach for hierarchical memory organization which is based on the notion of term subsumption. Based on this proposal an algorithm for memory organization and retrieval should be designed, implemented and tested.

  • Schmid, U. (1997). Programmieren durch analoges Schliessen. Kognitionswissenschaft, Sonderheft "Analoges Schliessen", 6 (3), 127-134.

  • Schmid, U., Sinha, U., and Wysotzki, F. (2001). Program reuse and abstraction by anti-unification. In G. Stumme et al.: Professionelles Wissensmanagement -- Erfahrungen und Visionen (pp. 183-185). Shaker. (German Workshop of Case-Based Reasoning (GWCBR2001) im Rahmen der WM 2001, 15-16 March, Baden-Baden.)

(2.1) Programming by Analogy: Retrieval and Memory Organization

In the context of the system IPAL, programming by analogy can be used in combination with planning: First, a plan for a small domain (say transporting three objects) is generated, then this plan is trannsformed into a finite term. The finite term is compared with recursive program schemes in memory and a generalized/recursive function which solves more general problems (say transporting n objects) is constructed by analogical transfer. Since planning an plan transformation are costly, it would be more efficient to use analogical transfer directly for the planning domains.

  • Toussaint, J.; Schmid, Ute; Wysotzki, F.: Using Recursive Control Rules in Planning. In: Arabnia, H. R. (Hrsg.): Proc. of ICAI'01 (ICAI'01 (Session "Learning and Adapting in AI Planning") Las Vegas, Nevada, June 25th-28th). Bd. II. Las Vegas: CSREA Press, 2001, S. 1012-1015.

(2.2) Planning by Analogy
Our approach to analogy is based on anti-unification. In contrast to transformational approaches to analogy, generalization learning occurs as a side-effect of analogical reasoning. To obtain evidence for the psychological plausibility of our approach, empirical studies should be performed.
Such studies could be in the domain of proportional analogies (e.g. the letter string domain), of predictive analogies (e.g. physics), or in analogical problem solving.

For proportional analogies, we have two designs in mind:
  • (a) Two different groups of subjects are acquainted with a sepcific analogy first. Thereby, generalized rule (i) should be learned by one group, generalized rule (ii) by the other group. In the following, different analogies with answer alternatives (i) and (ii) (being congruent with rule (i) resp. rule (ii)) are presented. We hope for a significant difference in saelection frequencies in dependence of the group.
  • (b) Subjects are giving several proportional analogies for solving. Afterwards, for a much larger set of analogies, they have to decide whether they have seen an analogy before or not. The analogies in the test set contain original problems, problems congruent with the assumed generalizations, and distractor problems. We hope to show, that there is no difference between recognition original and congruent problems. (This design corresponds to studies in the context of prototype acquistion.)
(2.3) Generalization-based Analogies (For psychology students!)

 


(3) Different Applications of Problem Solving and Learning

[Archive: Old topics file, 2003]