Notes from meeting with Will, 29-Jun (TC) …

• A discussion about how a PhD is training in research — i.e. how to become a researcher, and not necessarily about the end result
• Number of phases to a PhD — by the end of Phase 1, it is expected that the student:
  • state of the art 
  • understands the terminology and majors players in the field
  • major pieces of work
  • main journals and conferences
  • could talk to another expert on a peer-to-peer level — “comes up to speed” with field
• the student should have a rough idea on what they want to do ~8-9 months in

One approach on how to find item to study:

• write an aim
• work out the steps needed to achieve said aim
• if one of the steps is found to be a show-stopper, focus on solving that issue

Other points:

• read about Open Dynamics Engine (ODE)
• Jon to write up aim/objectives for current ideas, and pass to Will for critique
• Meeting details
  • Will and Jon to next talk Thu 27-Jul @ 10am
  • Jon to email Will weekly summaries for next two weeks, whilst Will is out of the country
  • At 27-Jul meeting, work out a suitable date for Jon to come down to Reading next (aiming for early August)

Thursday’s reading …

Today had a look through the Heppner and Grenander paper, A Stochastic Nonlinear Model for Coordinated Bird Flocks, to see what information they have presented. This paper is important as it’s the primary source for the PSO algorithms. Unlike Craig Reynolds’ paper, this publishes equations for home use, which is useful!

 Interestingly, did a quick literature search on Citeseer for this paper, and other than the recent Lubij(?sp) fuzzy logic paper, no one has really compared the Reynolds model and the Heppner model. Another important thing to note is the fact that no-one has actually defined what a flock is, from what I can tell, in the 3 papers: the results indicate that the simulations “appear” to flock — they can’t categorically state “this model is flocking”.

Had a quick read through the Wikipedia entry for Conway’s Game of Life (http://en.wikipedia.org/wiki/Conway%27s_Game_of_Life)*, which appears to have started off the field of cell automata. Related from what I can tell, as a number of applications I’ve come across (e.g. the Excel VisualBots from yesterday, NetLogo, etc) are mentioned in both flocking and ACO type discussions … I suppose flocking simulations is ultimately pixels on a grid, so the comparrison is relevant.

*starting to get exasperated with IE as it’s certainly no Firefox when it comes to blogging :( 

 Also read Fish in Schools (E. Shaw), Natural History, 1975, Vol 84, pp 40 – 46. Didn’t tell me much I didn’t already know, but interesting points about fish responding better when school neighbours are in certain areas of their vision; also, when a fish is in the middle of the school, it only has to concentrate on what it’s neighbours do — those on the outermost fringes need to concentrate on it’s neighbours and the world… interesting approach: surely those on the outermost edges of a school would expend more energy due to higher mental activitites?

Summary of Activites, Thurs 1-Jun-06 — Wed 29-Jun-06 …

Usefully I’ve spotted that I haven’t done one of these for a few weeks, so a quick blat through the last couple of weeks:

• Moved house, meaning I lost about 1 week, but gained backache !
• Got a 2D flocking simulation sort of working: the birds will now show the direction they are pointing in, and track where they’v been. They won’t stay *exactly* in the boundaries I want them to, and they always appear to fly to the left … so there are a few nuances that I need to iron out!!
• Reading: been looking into a few areas, including:
  • teamwork & leadership: can’t decide if this is of relevance or not, but looked a little into trust, leadership, etc. Trying to come to a conclusion in my head about the importance of trust and experience in flocks: is it required? Found a really interesting paper about Great Tits, Parus Major, who have the concept of status in flocks: highly interesting, as this opens up new ideas on how birds flock, and how they react/respond to individual agents of high status.
  • particle swarming/swarm intelligence: having a brief look through, to see what the field is all about. Still reading at the minute, but aiming to write something up in the form of a report

Meeting Agenda with Will, 29-Jun-06 @ 11 – 12pm (TC — Jon to call Will) …

• Summary of activities currently
  • reading ahead of preparing 2nd report (around Swarm Intelligence (mainly Particle Swarm Optimization, but some Ant Colony Optimization, plus a misc. section for anything else I come across)
  • reading through recently obtained inter-library loaned books and journals
• Query about what to do with thoughts for main thrust of research
  • (NB this could be a discussion best left until my next Reading visit, but anyway ..)
  • I’ve had a number of thoughts about what I could do in the main research phase, and I suppose my question is: what do I do with them? Would a structured literature research approach around each one be worthwhile, with the sample structure of:
    • what the idea is
    • what the literature says about this idea (done before, mentioned, etc)
    • what it could be useful for
      • (NB this would be over a few pages, <10, to have some sort of idea what’s out there)
  • I think a chat about what the main product of each of the sections of a PhD would be worthwhile (this could be what we discuss when I’m on campus next), so that I have a clearer idea in my head on what needs to be accomplished — i.e. some idea of what the “bigger picture” is and why

Notes from Wednesday …

We’re sort of back in action … on a temporary laptop, without normal applications, files, etc., which is a hindrance. Bit strange starting back up — what was I doing last? kind-of thing — but the blog is a great way of seeing where I was last time!

NB Found this site today when looking at PSO stuff: http://www.visualbots.com/index.htm which is a multi-agent simulator for Excel, enabling you to use VBA to control robots.

Read a bit about Ant Colony Optimization (ACO): it appears that when ants in the real world are looking for food, they wander around pretty randomly to begin with. When they find a food source, they walk between the food and their nest, laying down pheromones as they walk back and forth. These trails are sensed by other ants, meaning that they walk up and down them as well, reinforcning the trails. However the trails evaporate over time, so longer paths could evaporate, meaning that shorter paths tend to be reinforced more and more.

ACO appears to take this concept and put it into an algorithmic form, whereby ants by wandering around, finding routes to goals and reinforcing them, can find good solutions to problems. The evaporation of the paths are important though, as they ensure that the ants don’t get caught up in what are called local minima (unsure exactly what this means at the moment, but could have a guess). There is the whole topic of Optimisation http://en.wikipedia.org/wiki/Optimization_%28mathematics%29 which appears to be of relevance in this area, which I will have a quick skim through.

NB Came across this chap’s blog: http://pensive-pondering.blogspot.com/ — worth subscribing to?

Monday 26-June: we’re back !! (sort of…) …

Following a week off, we’re back in PhD business — kind of. Yesterday was pretty  much destroyed by a non-working laptop, meaning that IT woes meant I did nothing … there is also the issue about what data I’ve lost, if any. Gone to check …

Looks like all my files are in my backup — phew! Struggling with the adjustment back to early mornings though — overslept by 35min today, so reverting to hiding my alarm clock so I have to track it down in the morning.

Lots of journals I requested have arrived via email, so I’ve gone to battle with Adobe to see I can get them to print … 

Friday: Swarm Intelligence …

Slight change from yesterday’s idea of a PSO report: now shooting to deliver a Swarm Intelligence report instead. I’m thinking of keeping each part of it fairly terse at this point: I’m aiming to cover PSO, Ant Colony Optimization and Stochastic Diffusion Search (Note to self: after this report, probably a good idea to do a report on the learning styles (reinforcement, etc) as per your notes from earlier this week).

I’ll aim to complete each of the above in silo, then send onto Will for critque — hopefully this will ensure that I’m always writing/delivering something, to help move forwards, etc.

NB Saw some really interesting sites today:

http://swis.epfl.ch/People/Jim/Jim_Pugh_SIS2005.pdf

• an article about using PSO for unsupervised robot learning

http://ai-depot.com/Essay/SocialInsects-Swarm.html

• nice introduction to PSO and collective robotics

http://www.cs.ualberta.ca/~kube/research.html

• (could be a player …) came up with robots with the ability to collectively work together to push a light-box around (that’s too heavy for one individual to move on their own)

Thursday: back to reading …

Today started reading again. First up: what the coaching aspect of RoboCup is. Appears to be where you have a coach, and set teams: the coach watches the teams play, and then advises the team it controls on how to play better. The team is unknown to the coach, who is niéve: the coach hasn’t seen the teams play before.

Found a thesis on the topic here — this also mentions the predator-prey chase scenario you thought of the other day.

NB Reinforcement Learning (Sutton and Barto) can be found here.

Started looking at Particle Swarm Optimisation — I’m going to start working on a report of the field, as no doubt this will be of relevance to me. Here is a sample applet where “flocks” look to find the maximum height of a mountatin.

Really quick thought to self … do the concepts of experience, status and/or trust exist in PSO?

Also, betting that reading the sync book would be a good move, and that attractors will be a good topic to know about … could these topics be factored into the PSO report?

No updates today …

No updates today, as I’m preparing for an interview at 9am — eek! ;-)

Some notes from Wikipedia …

What I found …:

TRUST
http://en.wikipedia.org/wiki/Trust_%28sociology%29
Trust is the belief in the good character of one party, they are believed to seek to fulfil policies7, ethical8 codes, law9 and their previous promises.

In sociology17 (and psychology18) it refers to an open relationship19 between people, or between people and social institutions such as a corporation or government. It is the belief by one person that another’s motivations towards them are benevolent and honest, and that the other’s actions stem from a relevant level of competence.

In psychology, it is integral to the idea of social influence22: it is easier to influence or persuade someone who is trusting. The notion of trust is increasingly adopted to predict acceptance of behaviors by others, institutions23 (e.g. government agencies24) and objects such as machines25.

Research has been done on the notion of trust and its social implications; for instance the work of Barbara Misztal26 attempts to combine all notions of trust together. She points out three basic things that trust does in the lives of people: It makes social life predictable, it creates a sense of community27, and it makes it easier for people to work together.

Trust and power28 intuitively might seem contrasting concepts, but sometimes they overlap; see Stockholm syndrome29 and charismatic authority30.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~
CONFIDENCE
http://en.wikipedia.org/wiki/Confidence
Confidence is trust or faith that a person or thing is capable. Self-confidence is having confidence in oneself. Overconfidence is having too much confidence.

Losing confidence is no longer trusting in the ability to perform. It may be reasonable as the result of past failure to perform, or unreasonable, because one "just has a feeling" about something.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~
EXPERIENCE
http://en.wikipedia.org/wiki/Experience
Experience as a general concept comprises knowledge15 of or skill in or observation16 of some thing or some event gained through involvement in or exposure17 to that thing or event. The history of the word experience aligns it closely with the concept of experiment18.

The concept of experience generally refers to know-how19 or procedural knowledge20, rather than propositional knowledge21. Philosophers dub knowledge based on experience "empirical knowledge22" or "a posteriori knowledge".

The word "experience" may refer (somewhat ambiguously) both to mentally unprocessed immediately-perceived events as well as to the purported wisdom48 gained in subsequent49 reflection on those events or interpretation of them.

Most wisdom-experience accumulates over a period of time50, though one can also experience (and gain general wisdom-experience from) a single specific momentary event.

One may also differentiate51 between physical52, mental53, emotional54 and spiritual55 experience(s).