Topics

1

General Discussion / Hugin Researcher Algorithms (Reasons and Supporting Literature ???)

« on: April 25, 2011, 19:12:28  »

Hi,

I have extensively used Hugin Researcher in my research involving application of BN for Network Management tasks. I thank the Hugin team for developing such a professional software and also for providing the timely support and help.

Now, I am writing my PhD thesis and have to describe and cite the literature related to all the algorithms implemented in Hugin. If you can answer the following queries, it will boost the quality of my thesis.

1. Why were these particular algorithms chosen to be used in Hugin. Like EM for parameter learning, PC/NPC for structure learning and Lauritzen Speigelhalter Algorithm for inference? Even though there are other algorithms for doing the same task.

2. Have these algorithms which have been used in Hugin, compared for their accuracy, performance, speed and memory requirements with other contemporary algorithms existing in literature?

3. What supporting literature and arguments can I quote in my thesis to justify the use of them in my research work? Please provide such information with authentic references.

Hope to hear from you at your earliest possible convenience.

Regards,

Abul

2

FAQ / Hugin Researcher Algorithms (Reasons and Supporting Literature ???)

« on: April 23, 2011, 14:34:48  »

Hi,

I have extensively used Hugin Researcher in my research involving application of BN for Network Management tasks. I thank the Hugin team for developing such a professional software and also for providing the timely support and help.

Now, I am writing my PhD thesis and have to describe and cite the literature related to all the algorithms implemented in Hugin. If you can answer the following queries, it will boost the quality of my thesis.

1. Why were these particular algorithms chosen to be used in Hugin. Like EM for parameter learning, PC/NPC for structure learning and Lauritzen Speigelhalter Algorithm for inference? Even though there are other algorithms for doing the same task.

2. Have these algorithms which have been used in Hugin, compared for their accuracy, performance, speed and memory requirements with other contemporary algorithms existing in literature?

3. What supporting literature and arguments can I quote in my thesis to justify the use of them in my research work? Please provide such information with authentic references.

Hope to hear from you at your earliest possible convenience.

Regards,

Abul

3

HUGIN GUI Discussion / Analysis Wizard : Specific Algorithm?

« on: April 05, 2011, 12:50:50  »

Hi,

I just wanted to check what specific algorithm is used in the Analysis Wizard to calculate the Data Accuracy. As far as I understood, inference algorithms will be used for this purpose.

I am aware of that there are inference algorithms like Kim and Pearl’s Message Passing Algorithm
, Lauritzen Speigelhalter Algorithm etc.

Which of them are used in Hugin Researcher ver 7.3?

Your help will be highly appreciated.

Regards,

Abul

4

HUGIN GUI Discussion / Analysis Wizard : Predicted value?

« on: April 14, 2010, 19:04:07  »

While using the Analysis Wizard for testing the prediction accuracy of the BN model, it gives a confusion matrix in the "Data Accuracy" pane alongwith the error rate.

Also, it marks "red" for cases for which the state was predicted incorrect.

But it does not display the "erroneously" predicted state.

For example, assume a node A having 3 states low, med, high. If the actual state was low (as given in the test case), but it did not predict low, but something else, which led to an error. What was this "something else". Was it med or high?

I need this information because I want to plot the actual state and predicted state for a node of the BN.

Any help will be appreciated.

-Abul

5

HUGIN GUI Discussion / Monitor Windows

« on: March 29, 2010, 18:10:40  »

When I display the marginal probabilities in monitor windows, the order of the states in each of the nodes does not come the same. Like for example, I have 3 nodes (A,B,C) with three states (L,M,H). Then in one case the monitor window displays (LMH) and in another (HLM).

I knew how to make them the same order. But now I forgot. Please help me.

6

HUGIN GUI Discussion / How do I interpret negative BIC scores?

« on: March 25, 2010, 14:53:40  »

Dear Hugin Support,

The more I read about BIC on the internet, the more I get confused. Some say we have to choose the model with highest BIC score and some say lowest BIC score.

Which one is right?

Also, how do I interpret negative BIC scores (which I generally get in the Hugin models learned through EM algorithm)? Which one is better: -100 or -200? Also, which one is better: 100 or 200?

Please reply to clear this confusion.

Regards,

Abul

7

C++ / Structural Learning - through API

« on: August 21, 2009, 12:08:03  »

Hi,

Has anybody tried structural learning through the C++ API? I am not sure how to read the data file (1000 cases, 5 variables, 3 states each).

Is it required that the domain be initially defined with the name of the nodes? This is because, when I tried to parse a data file, it said that there is no node with name XXXXXX.

What is the way to use the function learnStructure()?

Please reply with a sample code    .

Regards,

Abul

8

C++ / How to discretise input data?

« on: June 12, 2009, 13:29:27  »

Hi,

I have a question with respect to structural learning through C++ API in Hugin Researcher 7.0 with MS VC++ 6.0

I am importing a set of 5 continuous variables from another application, which I would like to analyse (just like in the GUI which has this facility) and convert them to discrete levels (3 levels - low, med, high). Can this be done through C++ API or not? If yes, how?

Another question is -  Can sequential learning be done on Influence Diagrams or it has to be done on BN ONLY? (either by GUIs or APIs).

I would appreciate your suggestions.

9

HUGIN GUI Discussion / Testing BBN model

« on: April 24, 2009, 20:39:02  »

Hi,

I would request answers to the following questions please.

1. I monitored a system for 5 variables and recorded 1000 cases (1000 time instances). Continous variables (from the system) were modelled as Interval nodes in Hugin with 3 states each. Then, I used the learning wizard to build the BBN using the NPC algorithm. Also I ran the EM learning which computed the CPTs for me.

2. Now I would like to test the BBN (with another the set of data) which I have created. What are the tools in Hugin to do this?

3. Also I have converted my BBN to an ID for decision making? I would like to check how my decisions on the test data are going to be? Basically I want to make sure the ID is making right decisions on test data, as well.

4. Can BBN make predictions with respect to time? That is (if we assume a case to correspond to a time instance), can it estimate the marginal probabilities, say 10 cases in future?

5. How can we feedback our decisions and marginal probabilities back to the system to make any corrective actions in the system settings? Is this possible in Hugin?

I would highly appreciate your response.

Regards,

Abul

10

C++ / Integrating - Hugin and Opnet

« on: December 15, 2008, 18:06:53  »

Hi,

I would like to use the Hugin inference features in a communication network modelling software called Opnet Modeler 14.5. (http://www.opnet.com)

As far as my observation is concerned, Hugin API (C/C++) can be used in the Visual Studio environment (say Microsoft VC ++ 6.0). Anyone who has used or come across this kind of integration of Hugin API with other softwares?

Is this possible? What features are needed in the other software to perform this integration? 

I am running Hugin Lite 6.9 (GUI 6.9 build 6904, API 6.7) on Windows XP.

Any input or suggestions from this forum are welcome.

Regards,

Abul

11

General Discussion / Speed and scalability of adapdation algorithm

« on: September 21, 2008, 18:55:37  »

Hi,

The Hugin Engine uses the adaptation algorithm developed by Spiegelhalter & Lauritzen (1990).

I would like to know

1. What is the algorithm itself?
2. Any analysis done on it's speed (of belief propogation) and scalability (based on number of nodes)?

I would appreciate your response.

Regards,

Abul

12

General Discussion / Why Experience Tables?

« on: July 29, 2008, 17:41:36  »

Hi,

I am a new Hugin user and currently working with Hugin Lite 6.9.

I would like to know, once we learn the structure of the BN from the Structure Learning process and go to the stage of finding the CPTs for each of the nodes, why do we need the experience tables?

Why doesn't the software estimate the CPT values from the data file which we provide as the input? I mean the expereince table adds another uncertainity factor (based on the choice of the value from the designer).

Can somebody explain the reason behind the experience tables? Why is NOT the observed data enough for calculating the CPTs?

Regards,

Abul