Archive

Archive for December, 2009

Meeting with Davide 14 Dec 2009

December 14th, 2009 No comments

Had a meeting with Davide to discuss current research,

we talked about what I had discussed with Paddy, and Davide seemed to think this is effectively the same as the E-DTN position paper, and that we should pursue these ideas further.

We decided that in lieu of getting access to largert datasets, we should look at the ones we have allready.  I suggested that we use the Tom dataset that was collected over a week or so when Tom used the N95 I programmed – as a start this might help us to see what the data looks like. He suggested the following tasks:

  • Contact Eiko Yoneki to see if she has some datasets that we could use
    • UPDATE: She didn’t have any datasets and she is eager to get some herself…
  • Look at Grahams simulator, understand it, and extend it to include the ability for location data to be incorporated
  • Generate two graphs from the Tom data (which he called the TomStalker dataset):
    • Latitude vs Longitude vs Number devices seen (3d)
    • Frequency of device spotting vs time
  • Look into taking a statistics course at ucd, to learn the techniques of statisical analysis
  • Include him in correspondance about this

Meeting with Prag 9 Dec 2009

December 10th, 2009 No comments

Had a brief meeting with Prag Sharma, who described to me the sort of things that the clique group were doing.

I explained to him that I was interested in ways of analysing social networks in terms of movement patterns. He mentioned a few datasets that clique has some access to: Conrad and Fergul have access handset data from 6 million Nodes from a telephone network – IDIRO, but he did not know what the data included. Another dataset was NORON data, which is to do with financial fraud and included banking transaction data.

He suggested that a good person to talk to was Derek Green, who he thought was doing  similar work.

We saw Derek in is cube, and it seems he is looking more at social clustering, but we thought there might be some interesting overlap, so I will send him my position paper, and he will send me his recent presentation.

Prag suggested I check out the clique website.

Meeting with Paddy 7th Dec 2009

December 8th, 2009 No comments

Had a meeting with Paddy (Audio here)

Actions:

  • project/ experiment -find some location data extract and parse it,
    • Have to have a set of questions we are going to ask of the data
    • Looking for structures in the data – what are they?
  • extract soem temporal and strcutural patterns
  • Speak to prag (network clique) to find techniques on how to extract patterns from these networks.
  • Email Simon, Aaron and Adrian, to see if they know about, or have dealt with any large datasets which include accurate location information.
  • In 6 weeks present to a group of people about initial findings, and ideas for further research. but beforehand, keep in sync with Paddy, and make sure I don’t go off track. Keep a record in a wave about progress.
  • Also, investigate movement patterns.

Justifications for this:

Burst Situations – emergency situations – however movement patterns are different in this case.

Focus in on simple pragmatics – I want to be able to send free messages.

Pick social everyday applications that justify it – (parasitic networking?).

Need to focus on detailed stuff – get numbers,

Focus on movement patterns – what recurring patterns are in there – need to data mine it all.

How do we collect data, sufficiently large, and in constrained sets.

Padraig Cunningham group doing network analysis – We can use Paddy’s Social network to get help with things like statistics etc.

Extracitng movement patterns from everyday life – augemtnting with context information.

Set up experments

do blind first  – capture everything from individula, apply analysis just to movement informations plus simulation, then you take in to accoutn context information and see if it improves anything. Theres nothing to say that context is useful all the time. Perhaps we find it reduces the number of hops, this reduces failure rate.

Chapter 1 is the broad view – DTNs Human Movent – underlying model of data transmission. my hypothersis is that human moevemtn patterns will provide an underlying model for transmission of data on ad-hoc networks.

Next actions:

Think up and construct an experiment and do it – short term 3-4 week expt whewre we actually get somehting out of it – needs to be a small constrianed expriment, that takes into account movement and context.

Have to have a set of questions we are going to ask of the data:

Looking for structures in th data what are they?

couple of conversations: Prag Sharma – good talking point – i’ve got this data, I want to use thi

try to characterise the patterns – they might be structural patterns, or temporal –

Skype Meeting with Paddy 6 Dec 2009

December 7th, 2009 No comments

Had a skype meeting with paddy, and spoke about what I was going to do regards PhD (2009-12-07-time-15_24_48-Skype Call With Paddy)

  • I need to: Persuade Paddy that Natural Dynamic Networks idea don’t already exist in the literature
  • Convince him that that eperiments that I might construct are actually interesting for some reason – i.e. what is the point
  • Really nice to have 1 paragrah story line that we both agree tomorrow that we only change if the experiments tell us so, not just because something else comes along.

Need to be able to say – here’s my definition, here’s what it means, here’s what im gonna experiment on in the next 3-4 months.

Task is to come up with bullet points:

  • here’s my definition of NDN
  • hows its different from the way others people view networks?
  • If this is true, then we can do the following … A, B, C
  • if this is tru, then the experiments we can do are …

Forms the starting points of tomorrows meeting, that, we believe the first statement – that nobody has done this before.

Paddy will try to find some literature to find a few rocks to throw at me! so we can defend it.

If we can defend that – we can really question – we conjecture that we might be able to do the following.

Get arguments prepared for:

If we can do opportunistic routing – what do we get out of it, why is it better than what we have at the moment?