Hi
I am new to Genie, and I am trying to use it to model a group of persons.
I have designed the network , but Genie crashes as soon I press the F5 key
Could anybody tell me why this happens?
Thanks
I am using Genie version 2.0.4497.0
Genie crashes on this network ... why?
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
Re: Genie crashes on this network ... why?
The jointree generated by default for your network is quite large due to network structure. One of the cliques has more than 129 million entries, each entry is a double precision number (8 bytes). Effectively, approx. 1 gigabyte of contiguous memory is required to represent such clique. Since SMILE has less than perfect error handling, the std::bad_alloc exception is not caught and the program goes down.
You have two options if you want to proceed with this network:
1) mark a subset of nodes as targets - this can make jointree simpler
2) add DSL_REL_DECOMP_THRESHOLD property to the network with the value of 1 and switch to 'Relevance-based decomposition' algorithm.
You have two options if you want to proceed with this network:
1) mark a subset of nodes as targets - this can make jointree simpler
2) add DSL_REL_DECOMP_THRESHOLD property to the network with the value of 1 and switch to 'Relevance-based decomposition' algorithm.
Re: Genie crashes on this network ... why?
Thanks!
I have tried applying both your suggestions and now I am able to use "Update immediately" (Ctrl-F5) without Genie exploding immediately.
Unfortunately the network still explodes later when I do as follows in the main network:
- I select nodes n88 n87 n86 n77 n83 n78 n79 n80 n81 n82 as target nodes
- I use nodes n89 n84 n85 n76 n75 n70 n71 n72 n73 n74 to insert evidence by selecting in each one one of the three options
- when I select the value of the last one (in any order I do it) Genie explodes
Could you kindly explain what is happening?
Is there any way I could diagnose the problem by myself?
I have tried applying both your suggestions and now I am able to use "Update immediately" (Ctrl-F5) without Genie exploding immediately.
Unfortunately the network still explodes later when I do as follows in the main network:
- I select nodes n88 n87 n86 n77 n83 n78 n79 n80 n81 n82 as target nodes
- I use nodes n89 n84 n85 n76 n75 n70 n71 n72 n73 n74 to insert evidence by selecting in each one one of the three options
- when I select the value of the last one (in any order I do it) Genie explodes
Could you kindly explain what is happening?
Is there any way I could diagnose the problem by myself?
-
- Site Admin
- Posts: 1417
- Joined: Mon Nov 26, 2007 5:51 pm
Re: Genie crashes on this network ... why?
The network is too complex for clustering given the target and evidence sets (I assume you're running relevance decomposition with DSL_REL_DECOMP_THRESHOLD set to 1). You can consider one of the sampling algorithms, like EPIS.Could you kindly explain what is happening?
You already did by crashing the programIs there any way I could diagnose the problem by myself?
Re: Genie crashes on this network ... why?
Thanks!
it works
it works
Re: Genie crashes on this network ... why?
Quite courageous of you to form a circle in a BN! But seems it works
Some more things you could try (if you have the time to do so manually) is to define the nodes with noisy-MAX parameters, and to divorce them. That really works wonders sometimes
Some more things you could try (if you have the time to do so manually) is to define the nodes with noisy-MAX parameters, and to divorce them. That really works wonders sometimes
Re: Genie crashes on this network ... why?
Not exactly a circle ... the dependences are not circular even if the subnet network seems to be circular
I'll try your suggestion, thanks
I'll try your suggestion, thanks
Re: Genie crashes on this network ... why?
I have now able to successfully run the network with the exact Clustering algorithm.
To do that I have removed all its deterministic nodes, as they were used only as aliases (identic copies of other nodes) and were not actually needed.
To do that I have removed all its deterministic nodes, as they were used only as aliases (identic copies of other nodes) and were not actually needed.