Re: BNs from Decision Trees

Finn V. Jensen (fvj@cs.auc.dk)
Wed, 23 Sep 1998 09:35:15 +0200

Michael Mayo wrote:
>
> Here's a question for you all...
>
> If I have a decision tree with probabilities attached to each and
> every outcome arc and leaf nodes representing conclusions, are there
> any algorithms for inducing a bayesian network?
>
> Has anyone done any work on this? Would it be possible to induce a
> bayesian network with a single conclusion node where each value
> corresponds to a leaf node in the DT?
>
> MIke

If you have a symmetric decision tree, you can convert it to an
influence diagram. Yous may think of influence diagrams as BN's extended
with nodes for decisions and utilities.

Actually, influence diagrams were introduced as compact representations
of symmetric decision trees.

Some work has also been done on converting asymmetric decision scenarios
to IDs.

/Finn

Howard and Matheson(1981) Influence diagrams. Reprinted in "The
Principles snd Applications of Decision Analysis, vol 2 1984

Shachter (1986) Evaluating influence diagrams. Operations Reserach,
34(6) 871-882.

Shenoy 1992. Valuation-based systems for bayesian decision analysis.
Operations Research 40(3), 463-84.

In the latest UAI proceedings you can also find variuos papers on
ifluence diagrams.

/Finn