Staff Talk Schedule for CM50175
Research Project Preparation   2007

Projects

EE

• Duncan Allsopp's projects.  He isn't speaking, but wants interested people who are interested to come talk about it.
• Adrian N. Evans
• Research: image processing & analysis; image & video comprehension -- finding the one word in a thousand from a picture.
• segmentation, representation, recognition,
  
• SAR(adar) segmentation using local graulometries (texture)
• texture preserving speckle reduction
• cloud tracking (infrared) tracking deformable objects
• colour edge detection & segmentation
• Projects:
• Speckle reduction using granulometric sieves.
• Nonlinear texture segmentation
• Motion assisted video enhancement
  
• low quality video sequence (poor light or high compression)
• get stuff from a variety of frames & improve quality
  
• Paul Leonard has a link:
• Mostly programming, likes java, would consider others (c, c++, matlab).
• Artificial Life -- starts from existing code
• model aspects of life by siulating simple cells
• programming in java (OOD)
• http://www.bath.ac.uk/~eespil/GA/
• cells live in a 2D space (could extend to 3D but more costly)
• Cells use energy to tick & move
• cells gain energy by eating food at current position in space
• subregion space has food, but slower growth rate than cells need to survive so they must move.  
  
• asexual reproduction, (could extend w/ sex)
• functions to change state
• genes -- currently random generation of operators & connections, could combine (breeding, mutations)
• projects: implement sex, do plants & animals
• Music -- sound synthesis, algorithmic composition anything --- go talk to him!
• Internet applets, evolution of music
  
• Digital audio -- stuff like what he taught
• audio compression
  

Comp Sci

• Alessio Gugliemi
• Deep inference is a new methodology in proof theory, which is the discipline
  that studies mathematical proofs. Deep inference is about designing
  deductive systems and formalisms with excellent properties of analyticity,
  proof complexity and semantics of proofs. With deep inference, we express
  logics that elude the traditional methods of proof theory, in particular we
  represent modal logics and logics related to process algebras. Deep
  inference provides beautiful and simple deductive systems and formalisms.
  
  Sample projects at <http://alessio.guglielmi.name/res/cos/crt.html> and
  <http://moodle.bath.ac.uk/moodle5/mod/resource/view.php?id=11667>
• Computation steps trace of a process --- finite presentation, tends to be
• compressed, faithful
• Gugliemi's businesss is in logic, in fact in proof theory.
• combine axioms into a proof that proves your theory.
• a proof is like the trace of a process.
• input is the theorem, proof is the program
• proofs may be composed of modified versions of a simpler proof.
• can show relationships between the simple proofs & the full proofs.
• relationship between these compressed proofs and programs. 
  
• The mathematician uses intuition which is like the simpler proofs, the bigger proof unfolds everything, more like automated / compiled planning. 
  
• We know how to compress from the automated one to the mathematicians one.  There is a sophisticated (combinatorial) mathematics here.
• Gugliemi brings the anti-compression process closer to computer science.  e.g., finite numbers of iterations, no assumptions of infinite time:  deep inference.
• a combinatorial analysis of proofs for computer science logics.
• 8 year old theory
  
• MSc projects
• Maths, but not hard maths.  Don't need much  knowledge, but do need to be able to think about complex objects.
• People tend to understand this mathematics very quickly.
• Please read the web pages, papers.
• Theory projects:
• express an object that hasn't yet been expressed, e.g. another modal logic.
• after expressing it (if that was easy) can study its cut-elimination (how to connect the simple and the large proofs/programs).
• flexible, leads to a good publication!
  
• Practice projects (programming):
• have a proof editor for deep inference. Grape
• GRAPE automatically generates counter examples, very useful!
• crude, need to set it up.  No interface, needs to be made more usable.
  
• open source, doesn't require much maths, just need to improve the editor.
• do need to be able to read the papers.
  
• lots of contributions of various sizes, up to MSc how much work they do.
• Great community.
• Practical implications:
• bug free software!
• e.g. security:  need to prove protocols are safe, no information will be leaked.
• hard to prove because reasoning about infinite objects / languages.
• expects five years from now there will be major research in protocols & security --- big grants.
• eventually, hope to invite real programming languages people use that can be proved this way.
  
• James Davenport
• JHD uses mathematics to do computing, and vice versa. Specific areas of
  interest include computer algebra, cryptography and formal representation
  and presentation of mathematics (OpenMath/MathML).
• Here are the slides he talked with.
• Main project proposal was converting from open math mark up to the standard math markup language -- XMLs
  
• He also said he would do a project on cryptography if pursued.
  
• Russel Bradford has a projects web page too.
  
• interests:  cryptography, computer algebra, networks & sound
  
• log(xy) <> log x + log y if x & y are imaginary. (computer algebra, no projects in that)
• networks with canadians - making applications work with flakey internet by making a virtual internet that was flakey.
• cryptography --- traitor tracing (who gave away their key?)  Authentication -- knowing if you are buying from a real site.
• sound -- fourier transforms & analysis
• for sound talk to Prof. John Fitch
  
• projects
• networking
• graph layout:  large number of machines connected -- can you visualize it? 
  
• spring based to make it prettier / fit it out
• eigen vector method (for people who know what that is only -- very maths oriented.)
• visually-comprehensible network packet traffic
• chaffing & winnowing -- hide message in other random stuff (in plain sight).
  
• encryption
• new dutch advanced encription standard Rijndael
  
• implement it in several languages, compare & contrast results.
• implement several ways in one language
• digital watermarking -- recognize a photo or sound is yours without distorting it
• LEGO emulator -- build stuff with LEGO on the computer.
• 3d viewing, collision detection, etc.
• people try to do it but usually fail to do well
  
• Peter Hall
• I synthesise art from photographs, (semi)automatically. This invloves Computer Vision and Computer Graphics, occasionally calling on Machine Learning and human perception, as well as fine Art. It requires skills in mathematics and coding, along with an appreciation of the visual arts. Many possible titles, some to be suggested during the talk.
• Current research: perspective, perception & dynamic textures (clouds, water).
• Making cameras "see".
• Projects:
• Sudoku for the Visually Impaired
  Sudoku games printed in newspapers and magazines cannot be played by the visually impaired. The idea is to scan in a printed sudoku grid and present it on the computer in playable form.  The key point is  turning the scanned data into a grid that allows interaction, and which can be scaled in size.
  
  
• A Game of Chess
  Write a program you can play chess against. The key point is having the computer find the right move from amongst the many that exist.
  
  
• Draw a Tree
  Cartoon background are usually static – it is too expensive to make them move. Here is a possible solution. Take a video of a tree moving in the wind, use a computer to follow the leaves, repaint the video is some good looking way.
  
  
• Computer Vision
  If you like computer vision (make models from picture) , typically with the aim of making computer graphics later (make pictures from models), then talk to me about a project. Either one of your own, or I might be able to suggest one.
  
  
  
• Julian Padget, has everything on his new wiki.
  
• Research interests:  agents and institutions, semantic web services, multi agent systems
• project list
  
• Marina De Vos has also project pages.
• interests: answer set programming, knowledge representation and AI
• Presentation
  
• Answer Set Semantics   a <- b, not c.  : we know a if we know b is true but c is not .  Only assume truth if there is stong evidence.  Take the things we might want to believe and check them.  Try to figure out everything that is true!  (set of all rules)
  
• projects:
• Project 1: General Game Playing in ASP
• Project 2: Clima Competition Agent in ASP
• Project 3: Finite State Autonoma Generation
• Project 4: A Logic Programming Tutor
• Project 5: Playing Answer Set Programs as Games
• Project 6: A ccNUMA core for Platypus and / or an UltraSPARC T1 core
• Project 7: Profile and optimise smodels
• Project 8: A DNF based solver
• Joanna Bryson, see her projects link.
• John Collomose
• research interests: computer graphics & vision -- and particularly the convergence area between this, so using vision to construct graphics.
  
• Analysis & representation of visual information:
• non-photorealistic (artistic) rendering (NPR)
• content-based video retrieval
  
• real-time graphics & vision
• projects
• applying NPR to low-bandwidth video conferencing
• Automated piccaso - extend existing cubist software to not need marking up of the source image (combine with existing face detection software, for example).
• add new painting techniques to the set of NPR tools the department makes
• content based retrieval -including vision based localization
• sketch-based video retrieval
• gesture recognition
• video ? (can't read my writing!)
• physical world interaction -- using 2D bar codes with vision on a phone -- develop applications for this
  
• general mobile (phone) vision applications
  
• Phil Willis
• Is open to all kinds of graphics projects --- come to him with ideas.  Needed from you: software skills (probably C/C++)
• A few suggestions:
  
• An image compositor for film special effects etc. The project is to write a compositor with some novel techniques in it.  This has been done for years basbed on Porter & Duff (1984) but in the last couple years new ideas have come up to improve it.  E.g. make lighting effects.
  
• Virtual Physics VR only looks real, it doesn't act that way.  You can't solve all of physics (it would take too long) & collision stuff is largely already done, but there's a lot of things like flow of liquids or electricity that have not been done yet.  Examples -- water fountains, pipes, cups.
  
• Nicolai Vorobjov (easy to google)
• research : mathematical side of computer science -- all projects related to his interests
  
• example: examine geometric shapes using computers
• applications in graphics, but also more abstract
• may want to use them to compute!!
• how do we represent them such that a computer can handle them?
• finding alphabets that allow their description (algebraic geometry)
• topics for projects
  
• just some signposts, feel free to twist & turn into something you find more interesting & talk to him.
• computer linear algebra system -- like matlab, maple, mathematica.
  
• program for solving two-person games
  
• Beautiful Mind stuff!  John Nash
• represent moves with matricies, try to define what it means to be optimal -- what it means to play "well".
  
• useful for rock, paper scissors; chess; arms races; business modelling
  
• Robot motion planning
• optimal path planning around obstacles
  
• see full descriptions (last year's link).
  
• John (ffitch) Fitch
• interest: sound and music; compilers and software
• projects
• classifying sound (e.g. birds)
• software for composers, rehearsal, performance
• automated music, drum machine
• modelling physical instruments (only for mathematicians)
  
• sonification of data
• simple agent-based algorithmic composition
• Asterisk-based internet jukebox.
• Claire Willis and Derek Paddon talked about modelling and VR.
• They gave a handout.
• Alwyn Barry's projects.
• My primary interests are detailed on my website at:
  
  http://www.cs.bath.ac.uk/~amb/index.php/?page_id=3
  
  The main interest is in Genetics-based Machine learning, within which
  I am interested in techniques to evolve accurate minimal models of
  a problem space for classification or action selection purposes. The
  secondary interest is in Biologically-Inspired Computing, in which there
  are two directions ... the development of Spiking Neural Models to
  reproduce neural structures seen in nature and the development of
  models of cooperative animal behaviour.
  
  Since most students could not fully participate in these areas, some
  of the project ideas are rather more prosaic, only two being linked
  to current research ideas:
  
  
• Project Ideas:
  
• Implementation of Lion Hunting behaviour on 3D Robot Simulator
  
• Using stop codons in reproduction of populations of rules
•  Web text provenance creation and maintenance
•  Universal reliable PHP interface to XML databases
  
• Automatic translation of Openoffice DB applications to Web Interfaces
  
• general interests: machine learning, artificial life (parts of AI).  Applications: data mining.
• research projects:
  
• Visualization of GBML (views population dynamics of population-based machine learner.)
• Group behavior of simulated robot lions.
• Simulated wolf group behavior in MASON.
• Aquatic group predators (not in his list linked above)
• tuna & killer whales
• requires reading & complex simulation.
• Grammatical Evolution -- learn new programs autonomously - in python.
• applications:
• a better multiple choice question / automated test system
• a tool for generating web pages that do database queries - produce forms, validation & update codes.  s/b better than MS access.