trip report progress report thesis plan kevin pulo 2003-09-11
TRANSCRIPT
Trip ReportProgress Report
Thesis Plan
Kevin Pulo
2003-09-11
Overview
• Brisbane Trip Report– Griffith University– University of Queensland
• Thesis plan
• Recent progress
(in no particular order)
Brisbane Trip Report – Griffith Uni
• Visited the Software Quality Institute (SQI), headed by Prof. Geoff Dromey
• Software engineers
• Invented Design Behaviour Trees (DBTs) and Genetic Software Engineering (GSE)
DBTs
• A method for designing software systems by defining “behaviours” of components from their requirements
• These are then combined to form the overall behaviour of the system
• Can then be used to generate skeletal code• Build the system out of the requirements, not a
system which satisfies the requirements• Behaviours are naturally visually represented
using trees
Typical Component DBTs
Requirement-R1:When the user opens the door the light goes on
Requirement-R2:When the user closes the door the light goes off
Integrating Component DBTs
Requirement-R1:When the user opens the door the light goes on
Requirement-R2:When the user closes the door the light goes off
Integrating Component DBTs
Satisfies both R1 and R2
Overall DBTs
Very Large DBTs
User Navigation Data
• They have issues visualising large DBTs
• Installed my code on their machines– General code cleanup– Interface simplification– Animated undo/redo
• Logging facilities to record user navigation– Will be useful in devising quality measures– Will provide a benchmark suite / corpus for
running the measures
DBT Improvements
• Tip-over convention instead of inclusion layout (basically the same algorithm)
• Some other misc features
• Send updated version to Griffith researchers
Brisbane Trip Report - UQ
• Visited Advanced Computational Modelling Centre (ACMC) at UQ
• Bernard Pailthorpe, co-director (former director of Sydney Vislab, SDSC’s Vislab)
• Research mainly in Scientific Visualisation, collaborations with many wide fields (biotech, medical, psychology, physics, maths, chem, marine, etc)
• Visited ViSAC - Visualisation Laboratory
ViSAC photos
Thesis Outline(Working) Fancy Title: “Techniques for Structural
Focus + Context Display and Navigation”
1. Introduction, Background, etc– Started literature survey
2. Models, Measures and Techniques3. Application to Relational Data
i. Inclusion Treesii. Clustered Graphs
4. Case Study 1: Design Behaviour Trees (DBTs)5. Case Study 2: FADE clustered graphs6. Case Study 3: Citation networks7. Conclusion, etc
Data Sources - Trees
• Design Behaviour Trees – Data #1– Inclusion and tip-over layouts– Mostly done
• Philogenetic Trees – ???– Neither inclusion or tip-over seem appropriate– Problem is length of edge usually indicates
time– How to represent this in inclusion layout?
Data Sources – Clustered Graphs
• FADE clustered graphs – Data #2– Uses my work from first year (extending to Recursive
Voronoi Diagrams (RVDs))– Generally graphs of various views of software– Relatively sparse– From Aaron Quigley’s thesis or use SE tools to
generate my own – or maybe even arbitrary graphs
• Citation networks – Data #3– From any/all of Citeseer, Web-of-Science, IEEE,
ACM, etc– Clustered according to hierarchical topic (where
available)– Relatively dense
Edge Routing
• Research possible algorithms
• Attempt implementation of simple/naïve one first
• Attempt harder ones only if simple one doesn’t suffice
• Develop edge animations– Topology changes are hard
Evaluation – Empirical Measures
• Devise 5-7 good measures of the quality of a Smooth Structural Zooming technique– Eg: number of different animation directions,
number of objects moving concurrently, amount of overlap between objects, etc
• Assertion is that these measures are a good representation of reality
• Apply the measures to a corpus of test data
Test Data
• Require test data of both graphs/trees AND navigation through them
• Hence the Griffith usage data
• Other possibilities:– Generated (random)– Hand-constructed specific cases (eg.
best/worst, “typical”)
Evaluation – User Experiments
• Devise tasks to test some hypothesis
• Considering not doing them
• Require a lot of time and work:– Ethics approval– Logistics– Many unknown details:
• Hypothesis = ?• Task(s) = ?• Measurements (eg. time, accuracy)
Sep 03
Griffith code
Deliverables: Tip-over + misc code
Due: Sept 26 (2 weeks)
Design Experiments (?)
Deliverables: Ethics approval request
Due: Sept 26 (2 weeks)
Oct 03
Jewelry Box SSZ
Deliverables: Code for SSZ of JB layout (incl. force-scan alg)
Due: 10 Oct (2 weeks)
Obtain Preliminary Clustered Graph Data
Deliverables: One small clustered graph from data #2 or #3
Due: 3 Oct (1 week)
Obtain full data #2 (FADE)
Deliverables: Dataset of > 10 graphs, varying sizes
Due: 24 Oct (2 weeks)
Nov 03
Research and implement routing algorithms
Deliverables: Code for edge routing and edge animation
Due: 28 Nov (7 weeks)
Collect Griffith Usage Data
Deliverables: Usage data + code to use it as input + devise measures
Due: 14 Nov (3 weeks)
Obtain data #3 (Citation networks)
Deliverables: Dataset of > 10 graphs, varying sizes
Due: 28 Nov (2 weeks)
Dec 03
Apply SSZ methods to data #2
Deliverables: Code operating on FADE graphs
Due: 12 Dec (2 weeks)
Write thesis chapters 1 - 3
Due: 2 Jan (5 weeks)
Apply measures to data #1 and #2
Deliverables: Results of measures
Due: 26 Dec (2 weeks)
Jan 04
Develop experiments (?)
Deliverables: Code, data, instructions for experiments
Due: 23 Jan (3 weeks)
Apply method to #3 data
Deliverables: Code operating on citation networks
Due: 23 Jan (3 weeks)
Feb 04
Carry out experiments (?)
Deliverables: Experiments done
Due: 20 Feb (3 weeks)
Write rest of thesis
Due: 26 Mar (9 weeks)
Mar 04
Analyse experiment data (?)
Deliverables: Results graphs, conclusions
Due: 12 Mar (3 weeks)
Backups
• Thesis/PhD work– ~ 900 Mb– 3 machines: Uni, home, laptop– Synchronised using rsync and rdiff-backup– Fast transfers, incremental backups
• Alternatives include rsync, file-unison, cvs, etc
– Also weekly CDs...?
• Laptop– Daily/weekly rdiff-backup of majority of system
(20 Gb) at uni and home