Is there anything comparable to MarcoPolo for those of us stuck with
the other operating system? In case there's not, I've started working
on some Python code to do the evidence sources (have working sources
for IP addresses, attached monitors, active network links, running
applications, and time of day), but I'm more than happy to just use
someone else's program.
On Sat, Jul 26, 2008 at 1:01 PM, Mike Renfro <mwren...@gmail.com> wrote: > Is there anything comparable to MarcoPolo for those of us stuck with > the other operating system? In case there's not, I've started working > on some Python code to do the evidence sources (have working sources > for IP addresses, attached monitors, active network links, running > applications, and time of day), but I'm more than happy to just use > someone else's program.
I believe there was an effort to write something like that, initially targeted at Gnome/Linux, called GeoClue (http://www.freedesktop.org/wiki/Software/GeoClue). It might be good to contribute your effort there, or at least see what they've got going on.
Looks like GeoClue is just focused on providing physical location to
other applications. So I'm continuing on with my poor imitation of
MarcoPolo. Proof-of-concept Python code is posted at [1].
On Sat, Jul 26, 2008 at 3:01 PM, Mike Renfro <mwren...@gmail.com> wrote: > Looks like GeoClue is just focused on providing physical location to > other applications. So I'm continuing on with my poor imitation of > MarcoPolo. Proof-of-concept Python code is posted at [1].
Your confidence combination system is different to MarcoPolo's, incidentally. Instead of summing the confidences, MarcoPolo multiplies their complement (1-confidence) and then takes the complement of the product. In your Python code, it would amount to changing lines to look like:
contexts[ruleContext] *= 1 - ruleStrength ... for context in contexts: print context, 1 - contexts[ruleContext]
Forgive me if I'm intruding here with some of my own work, but over the past several months, I've been working on a system similar to MarcoPolo called Pennyworth:
It's an open-source application that currently runs on the Mac and uses machine learners to learn and predict changes in user activity, location, and social context.
I bring this up because it fills a very similar role to MarcoPolo and this week, I began working on the Windows port of the Pennyworth system. I have a couple of sensors (evidence sources) running on the Windows .Net framework and I am aiming to have a complete beta by the end of August.
If you're interested in this Windows port and would like to be kept up to date on its development, please e-mail me off-list.
> Is there anything comparable to MarcoPolo for those of us stuck with > the other operating system? In case there's not, I've started working > on some Python code to do the evidence sources (have working sources > for IP addresses, attached monitors, active network links, running > applications, and time of day), but I'm more than happy to just use > someone else's program.
On Sat, Jul 26, 2008 at 11:50 PM, Chris Karr <cjk...@gmail.com> wrote: > Forgive me if I'm intruding here with some of my own work, but over > the past several months, I've been working on a system similar to > MarcoPolo called Pennyworth:
By all means, go for it. As should be obvious from the fact that MarcoPolo is free (as in speech) software, I'm interested in better software overall, not just my own project! I'm always happy for folk to discuss other related projects, especially their own.
On Jul 26, 1:53 am, "David Symonds" <dsymo...@gmail.com> wrote:
> Your confidence combination system is different to MarcoPolo's,
> incidentally. Instead of summing the confidences, MarcoPolo multiplies
> their complement (1-confidence) and then takes the complement of the
> product. In your Python code, it would amount to changing lines to
> look like:
> contexts[ruleContext] *= 1 - ruleStrength
> ...
> for context in contexts:
> print context, 1 - contexts[ruleContext]
New and improved version is up with these changes, and others (CIDR
addressing for IP addresses, time of day rules, miscellaneous code
cleanup).
|
