Microsoft Visual C++ Runtime Error

The front end.
Post Reply
will.s.barnett
Posts: 3
Joined: Sun Apr 11, 2010 8:21 pm
Location: Bozeman, MT

Microsoft Visual C++ Runtime Error

Post by will.s.barnett »

Hello, I am new to the forum. I"m a graduate student working on belief networks and their application to ecology. I went through all of the tutorials yesterday, and am in the process of trying to learn the parameters for a belief network I have setup.

I keep running into a runtime error, though. I noticed in a previous post that this might be related to the amount of available RAM on my computer. I'm working on a laptop with 1 GB of RAM, and no other applications running. Is this not sufficient?

I am happy to attach the network and the discretized dataset, if necessary. Thanks in advance. I'm really enjoying the GeNIe program!
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Re: Microsoft Visual C++ Runtime Error

Post by shooltz[BayesFusion] »

will.s.barnett wrote:I'm working on a laptop with 1 GB of RAM, and no other applications running. Is this not sufficient?
That entirely depends on the network and the evidence set.
I am happy to attach the network and the discretized dataset, if necessary.
Yes, post the files here.
will.s.barnett
Posts: 3
Joined: Sun Apr 11, 2010 8:21 pm
Location: Bozeman, MT

Post by will.s.barnett »

Thanks for taking the time to look at this. The network and dataset are attached. I also added a comment that explains the correspondence between the nodes and the column names in the data.
Attachments
bn07_discretized.zip
There are several variables in the dataset that do not have nodes. In addition, the node identified as veg corresponds to "veg07", "WaterPrev" corresponds to dat$X06Wk1, and "Water" corresponds to dat$X07wk1.
(1.73 KiB) Downloaded 338 times
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Post by shooltz[BayesFusion] »

I can reproduce the crash on the machine with 2GB. It's caused by excessive memory allocations during EM - we're checking if this can be fixed.

I was able to convert the nodes to plain CPT and run EM successfully. However, conversion back to noisy will be very long.
will.s.barnett
Posts: 3
Joined: Sun Apr 11, 2010 8:21 pm
Location: Bozeman, MT

Post by will.s.barnett »

Thanks for checking into the error. Maybe I don't understand the nuances of Noisy-Max nodes very well, but shouldn't they simplify computation? I thought that the conditional probability table of the child node (in this case, Milfoil) would be huge if I left it as a "Chance-General" node, and would have much fewer entries with the Noisy-Max designation.
shooltz[BayesFusion]
Site Admin
Posts: 1417
Joined: Mon Nov 26, 2007 5:51 pm

Post by shooltz[BayesFusion] »

will.s.barnett wrote:Maybe I don't understand the nuances of Noisy-Max nodes very well, but shouldn't they simplify computation?
They simplify modelling. Computation requires separate code paths activated when noisy nodes are present, or simply uses expanded CPT.

The crash you've found is in the noisy-specific part of EM.
Post Reply