[prev in list] [next in list] [prev in thread] [next in thread]
List: wekalist
Subject: Re: [Wekalist] Bayesian networks
From: Remco Bouckaert <rrb () xm ! co ! nz>
Date: 2013-01-31 23:55:03
Message-ID: 1359676503.2347.11.camel () computer
[Download RAW message or body]
On Tue, 2013-01-29 at 11:29 +0530, sandeep varma wrote:
> i am using bayesian networks to construct dag network using simple
> estimator and ICS search algorithm with markov blanket option set to
> on.
>
> i have set the alpha parameter to 0.5. however as per the weka
> documentation when i setting the parameter to zero for maximum
> likelihoods it is showing gammaa error. so i have the set the alpha
> parameter to 0.0001. am i proceeding in the wright way in determining
> the likelihoods?
The documentation refers to estimation of the conditional probability
tables. These are the probabilities (not likelihoods) of a node in the
network having a value, given the values of all its parents. Maximum
likelihood is one method for estimating these probabilities. But if your
data does not contain the combination of node-value and parent-values,
the maximum likelihood estimate will set the conditional probability to
zero. This leads to problems, for example when that particular
combination is present in the test set.
> Also one more clarification required.
>
> the weka tool shows the calculation of the conditional probabilities
> using the formula shown in the below sanpshot. i could understand the
> numerator portion. the denominator portion is quite confusing?
>
> what does it signify exactly with respect to the dataset. what values
> should we consider in the denominator given a dataset
N_ij = \sum_k N_ijk,
and
N_ij' = \sum_k N_ijk'
Hope this clarifies things.
Remco
_______________________________________________
Wekalist mailing list
Send posts to: Wekalist@list.scms.waikato.ac.nz
List info and subscription status: https://list.scms.waikato.ac.nz/mailman/=
listinfo/wekalist
List etiquette: http://www.cs.waikato.ac.nz/~ml/weka/mailinglist_etiquette.=
html
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic