Author Topic: Limited-Memory Influence Diagrams vs. Traditional Influence Diagrams  (Read 29539 times)

Offline Anders L Madsen

  • HUGIN Expert
  • Hero Member
  • *****
  • Posts: 2295
    • View Profile
There are important differences related to influence diagrams between HUGIN 6.9 and HUGIN 7.x.

In HUGIN 7.0 we introduced "LIMIDs" (Lauritzen & Nilsson'01). The support for LIMIDs changes the semantics of the information arcs in the diagram and the solution algorithm.

In the traditional influence diagram (HUGIN 6.9) we assume the decision maker to be non-forgetting and use the Jensen, Jensen & Dittmer (94) algorithm to solve the diagram. This implies that the decision maker is assumed to recall all past observations and decisions. By this assumption some information arcs are assumed present. E.g., for the last decision all observations prior to the first decision are assumed observed due to not-forgetting. The diagram is solved by solving for each decision in reverse time ordering.

In the LIMID, all information arcs should be explicitly drawn in the diagram. There is no assumption about perfect recall (non-forgetting). Thus, an information arc from node X into the first decision does not imply that we assume an information arc from node X into any later decision to be implicitly present. Hence, we may model that the decision maker is forgetful.

Also, the solution algorithm is changed. The solution algorithm for LIMIDs is Single Policy Updating where we iteratively solve for each decision (in reverse time order, if an ordering is present). The user has to press the "SPU" button in the toolbar to run Single Policy Updating after compiling the network in order compute updated decision policies.

To change your model into a traditional influence diagram (if you want), you will have to add information arcs to the diagram.
« Last Edit: November 11, 2010, 17:29:23 by Anders L Madsen »
HUGIN EXPERT A/S

Offline MarcoGV

  • Newbie
  • *
  • Posts: 15
    • View Profile
Thank you for the explanation.  It would be nice to have (links to) examples of LIMIDs, especially ones that emphasize the difference between LIMIDs and perfect recall IDs.  An explanation of the various policy buttons in Hugin would be useful also.

Offline Anders L Madsen

  • HUGIN Expert
  • Hero Member
  • *****
  • Posts: 2295
    • View Profile
Dear Macro,

The policy buttons in HUGIN GUI are described in the online version of the HUGIN GUI html pages. You will find links to the descriptions under the description of SPU:
http://download.hugin.com/webdocs/manuals/Htmlhelp/descr_update_policies.html

The paper by Lauritzen & Nilsson on LIMIDs describes the Breeding Pigs example. We have made a description of the model and a simple interface here:
http://demo.hugin.com/index.php/Breeding_Pigs

There are not yet many LIMIDs described in the literature. This example uses LIMIDs to implement different mitigation actions that are not ordered:
http://demo.hugin.com/index.php/Catchment_phosphorous_run-off_network,_Morsa_Norway
HUGIN EXPERT A/S

Offline MarcoGV

  • Newbie
  • *
  • Posts: 15
    • View Profile
Re: Limited-Memory Influence Diagrams vs. Traditional Influence Diagrams
« Reply #3 on: December 05, 2013, 22:04:36 »
Dear Anders,

Are there any new examples of LIMIDs that have been published since your previous post of April 13, 2012?

Thank you and best regards,

 Marco

Offline Anders L Madsen

  • HUGIN Expert
  • Hero Member
  • *****
  • Posts: 2295
    • View Profile
Re: Limited-Memory Influence Diagrams vs. Traditional Influence Diagrams
« Reply #4 on: December 20, 2013, 12:19:11 »
Dear Marco,

Quote
Are there any new examples of LIMIDs that have been published since your previous post of April 13, 2012?

Not that I can recall. I would be very interested in hearing about this too.

Best regards
Anders
HUGIN EXPERT A/S