Help with influence diagram

The front end.
Post Reply
orochisob
Posts: 7
Joined: Wed May 25, 2022 11:42 am

Help with influence diagram

Post by orochisob »

Hi,
I am developing an influence diagram for risk assessment. I have attached my model here. The initial plan was that the nodes in the Risk control measures (RCM) layer should be modelled with decision nodes. These RCM denotes that the decision-makers can make decisions if they want to implement the risk control measures or not based on the observed utility. However, the model crashes when using the decision nodes as there are 19 RCM nodes. please let me know if there is any way to solve this issue. Now I am using the deterministic nodes instead of decision nodes which does not crash the model but I need to change the decisions to implement or not manually and compare all alternatives.
Thank you in advance.
Attachments
example.xdsl
(34.25 KiB) Downloaded 6 times
marek [BayesFusion]
Site Admin
Posts: 392
Joined: Tue Dec 11, 2007 4:24 pm

Re: Help with influence diagram

Post by marek [BayesFusion] »

Your model is quite impressive (in terms of size and complexity). I am not sure you can solve it using our default ID algorithm, as it will involve calculating the expected utilities over 2^19 (524,288) different policies. This is, of course, when you change back the RCM* nodes to decision nodes. Are you sure you want this :-)? May I propose that you instead solve the diagram using the "Find Best Policy" algorithm rather than "Policy Evaluation". The "Policy Evaluation" algorithm will derive the expected utilities of all possible policies and there are over half a million of them. "Find Best Policy" will step by step tell you what to do next. It seems to me that this is what you want (or should want :-)).
I hope this helps,

Marek
orochisob
Posts: 7
Joined: Wed May 25, 2022 11:42 am

Re: Help with influence diagram

Post by orochisob »

Hi Marek, Thank you for your suggestion. I tried it but it says "Cannot solve this network using SHACHTER's algorithm. Error while merging utility nodes (too many parents)". As the error says, I tried it with a few utility nodes but couldn't solve the error.
marek [BayesFusion]
Site Admin
Posts: 392
Joined: Tue Dec 11, 2007 4:24 pm

Re: Help with influence diagram

Post by marek [BayesFusion] »

Clearly, your influence diagram is too complex. Please keep in mind that probabilistic inference is worst case NP-hard and it is not that difficult to construct models that are not computable. Quite likely if a model cannot be solved by SMILE/GeNIe, you cannot solve it with any software on Earth. The easiest thing will be to make it simpler.
I hope this helps,

Marek
orochisob
Posts: 7
Joined: Wed May 25, 2022 11:42 am

Re: Help with influence diagram

Post by orochisob »

Hi again,
Yes. I think I will have to work on simplifying the model and maybe just stick with the deterministic nodes instead of decision nodes as it still gets the work done. Thank you!
Post Reply