[melbourne-pug] mirror app [Was: Project time?]

[Mike Dewhirst]

On 5/02/2010 3:36pm, Irae Hueck Costa wrote:
>
>
> How about a little brainstorming and then we meet at some McDonalds?
> Here an idea:
> Feed the data into an artificial neural network (and make the found
> patterns usefull for the everyday use?).

Let's accept that the data sets currently available are less than great. 
If we assume (invent) some good data then we can write a winning 
application.

We need to look past current datasets. If what we do is compelling 
enough it will eventually force the proper collection of useful data 
just so our application can reveal insights otherwise unattainable.

I had discussions in 2007 with the Vic government about exactly this 
requirement. We agreed then that the objective should be to correlate 
information so that it is possible for policy makers to adjust something 
(a law, a tax, a handout, whatever) with a predicted outcome and for the 
system to demonstrate success or failure down the track.

For example, let's say water consumption is particularly low in a 
partularly crime-ridden area. A policy maker might ease watering 
restrictions there with the expected outcome of greener gardens hoping 
community morale might pick up and the crime rate may drop. BTW if it 
worked it might have nothing to do with morale but rather the increase 
in casual employment for young gardeners.

All of that is just ordinary. What we really need is the equivalent of 
time-lapse photography and data collected in lots of (de-identified) 
streams which are tied to geographic locations.

Patterns of information in the environment indicate trends in society. 
We just need to see them in context to draw the correct insights. 
Remember New York's zero tolerance of graffiti co-inciding with a drop 
in homicides?

So, I think our competition entry should be an API for a data 
visualisation app. Or a wrapper for an existing data visualisation tool. 
I want to see "time-lapse" mountains and valleys flowing across the 
screen so our policy makers can spot connections, freeze the frame, 
drill down and ponder. A mirror on society.

We should be able to assume the existence of the NBN and widespread VPN 
access to databases hither and yon. For the competition we can 
manufacture whatever data we want for demonstration (actually testing) 
purposes.

Mike



> I dont think that this will give better results than other stochastic
> methods, but it sounds fancy.
>
> Here some shorter python specific introduction to neural networks:
> http://www.ibm.com/developerworks/library/l-neural/
>
> ps:
> Maybe we should use some other way to communicate about this project so
> we don`t flood the email incomming box of people who are not interested.
>
>  > From: r1chardj0n3s at gmail.com
>  > Date: Thu, 4 Feb 2010 10:05:44 +1100
>  > To: melbourne-pug at python.org
>  > Subject: [melbourne-pug] Project time?
>  >
>  > Hi all,
>  >
>  > So this competition was just announced:
>  >
>  > http://www.premier.vic.gov.au/component/content/article/9300.html
>  >
>  > The short story being:
>  >
>  > 1. take 50 datasets like
>  > • Maps of local hospitals (http://www.health.vic.gov.au/maps/index.htm)
>  > • Crime statistics (http://www.vicpolicenews.com.au/myplace.html)
>  > • Water resources data (http://www.vicwaterdata.net/)
>  > 2. crank up a Pinax website (allowing us to focus on point #3)
>  > 3. present information from #1 in "interesting" ways
>  > 4. profit!
>  >
>  > So, who's interested?
>  >
>  >
>  > Richard
>  >
>  > _______________________________________________
>  > melbourne-pug mailing list
>  > melbourne-pug at python.org
>  > http://mail.python.org/mailman/listinfo/melbourne-pug
>
> ------------------------------------------------------------------------
> Quer navegar na Internet sem medo? Instale grátis o Internet Explorer 8.
> <http://go.microsoft.com/?linkid=9707132>
>
>
>
> _______________________________________________
> melbourne-pug mailing list
> melbourne-pug at python.org
> http://mail.python.org/mailman/listinfo/melbourne-pug