A Visual

1
A Visual Interactive Computability Course
Emphasizing Breadth of Automata

• Rakesh Verma
• Computer Science Department
• University of Houston
• http//www.cs.uh.edu/rmverma

2
Outline

• The problem
• Proposed solution
• Teaching experience summary
• Conclusions future work

3
The Problem

• Learning and teaching Automata Theory are
  challenging tasks
• Abstract, difficult material
• Considerable creativity required
• Slow feedback in traditional format
• Student perceptions make it worse
• Dated material, of little use
• Dry at best and boring/frustrating at worst

4
Challenges Specific to UH

• Large, urban institution with a mature and
  diverse student population
• Some loss of mathematical knowledge
• Significant number of working students
• In-school tutorial sessions difficult

5
Goals

• Enhance learning and learning experience by
  increasing visualization, interaction, and
  faster feedback
• Expose students to
• A variety of automata
• Applications, especially recent ones
• Historical background and development

6
Proposed Solution

• Complete revision of course contents and teaching
  with
• integration of enhanced JFLAP (Duke U.) and LRR
  (UH)
• variety of automata, e.g., tree and DAG automata
• recent and classical applications of automata

7
Solution (contd.)

• We have developed
• Power point slides for lectures including JFLAP
  animations (LRR animations in F05)
• New problem sets requiring use of these tools
• Notes on tree automata and DAG automata
• Web site http//www2.cs.uh.edu/rmverma/3340new/h
  tml/index.html

8
LRR and its advantages

• LRR input is a set of rules and an expression
• Each rule is l ? r, where l and r are expressions
  with variables and operators
• Example x 0 ? x and x -x ? 0
• Applies rules via matching to given expression
• LRR can implement any computation
• A Turing machine can be encoded via just one rule!

9
Visualizing Tree Automata

• Instead of developing another package we use LRR
  with RuleMaker a graphical interface
• DFA example
• a(q0) ? q0 a(q1) ? q1
• b(q0) ? q1 b(q1) ? q0
• Final(q0) ? true Final(q1) ? false
• Final(b(a(q0))?

10
Examples (contd.)

• NFA can be simulated as below
• a(a(q1)) ? q2 (string) q0 ? q1 (epsilon)
• Tree automaton example
• true ? q0 false ? q1
• and(q0, q0) ? q0 and(q0, q1) ? q1
• and(q1, q0) ? q1 and(q1, q1) ? q1

11
UH Enhancements to JFLAP

• Debug feature students can generate strings
  at random and find the results for all the
  strings
• Equivalence checking for DFAs (this is now part
  of JFLAP version 4.0)

12
Preliminary Results

• Students participating in special problems and
  projects on automata

13
Prel. Results (contd.)

• Positive feedback on surveys regarding
  JFLAP/JFLAP and web-site materials
• Students from this course showed improved
  performance on subsequent compilers course
  compared to traditional versions taught by other
  faculty

14
Conclusions

• Automata theory can be made interesting and
  easier
• Visualization is helpful
• Interaction is key during lectures and outside
  class

15
Future Work

• Rigorous comparison with traditional automata
  courses
• Further development of materials on breadth and
  applications of automata
• A drastic revision of course contents

16
Acknowledgements

• Students
• Saquib Hakim (JFLAP and slides)
• Mohammad Anwar (web site)
• Pavan Podila and Patrick Sharkey (RuleMaker)
• James Thigpen (integration with LRR)
• Sponsor National Science Foundation
• The students who have enjoyed/suffered through my
  course!

17
Thank You!

• ?s or !s
• For more info http//www2.cs.uh.edu/rmverma/3340
  new/html/index.html