PDA

View Full Version : Long shot...Any experts in Bayesian Belief Networks?


bgoodheart
07-18-2012, 01:48 PM
Long story, but I'd love to pick someone's brain on Bayesian Belief Networks for probabilistic risk assessment and causal modeling. Beer (or other beverage of your choice) in exchange for ideas. Anybody with experience? Shoot me a PM, let's talk!

BJ

Red_Chili
07-18-2012, 02:17 PM
Sounds like it would be interesting to just sit and listen over a beer. Given sufficient beer I might begin to believe I was following the conversation.

DaveInDenver
07-18-2012, 02:42 PM
Wow, no kidding Bill, enough to be a nuisance... It's been a while since I sat through Dr. Papantoni's probabilistic and causal analysis class so it's rusty. Last time I even really reviewed it was for the P.E. exam and even then it was at a survey level since I wasn't taking the software engineering test.

I did do some Bayesian estimation on a SDR (modem DSP channel optimization) we built a couple of years ago, but honestly the PhD derived a lot of the heavy duty equations for we non-graduate degree'd peons.

So data mining, right?

bgoodheart
07-18-2012, 04:04 PM
Not so much data mining, per se. Bayesian statistics is a paradigm - the corollary to the frequentist mindset, which is the one most of us are familiar with (p-values, confidence intervals, all that jazz). You can certainly interpret data mining in a Bayesian framework, but most of what I'm doing has too much uncertainty for that (not enough hard data).

I'm building a predictive model via a Bayesian Belief network - really just a Markov chain that uses Monte Carlo simulation in the data generation phase and parent-child nodes to represent causal relationships - that should evaluate runway safety despite the rarity of events (runway incursions) and a lot of uncertainty :blah:. It sounds ugly when I try to type it here, but it is a really elegant modeling technique when you get down to it.

Like I said, long story, but pretty interesting stuff. I just need to pick the brain of someone who has done this on a large-scale model, and given the wide variety of folks in the club, I figured it's worth a shot posting here!

BJ

baja1d
07-19-2012, 08:11 AM
Long story, but I'd love to pick someone's brain on Bayesian Belief Networks for probabilistic risk assessment and causal modeling. Beer (or other beverage of your choice) in exchange for ideas. Anybody with experience? Shoot me a PM, let's talk!

BJ

You lost me @ "probabilistic". Must go weld something!!!