Logo name
Personal tools

Naive Bayes block

From Piki

  • Currently5.00/5
Jump to: navigation, search

The naive Bayes classifier block (also known as Idiot's Bayes) is a simple probabilistic classifier. It is based on a probability model that is based on highly questionable independence assumptions. The basic model assumes that there is no correlation between the input features.

Naive Bayes
Image:Naive bayes icon.jpg
Input ports1
Output ports1
Interactive GUINo

Despite its simple design and questionable assumptions naive Bayes classifiers can perform beyond any reasonable expectation. Since it is extremely fast to train and use, in many cases it is in practice superior to many more advanced methods.



The naive Bayes is a classifier that maps a number of input features to a set of labels (classes).

The component has two ports, one (top) that takes the input signal and the other (bottom) that takes the desired classification. The signal on the desired classification port has to have nominal enumeration (one feature per class, with 1 marking membership and -1 marking non-membership. This is only needed during training. With adaptation turned off, any data can be sent in on the desired classification port as it won't be used).

For more information on nominal encoding see details in tutorial 2.


p(C |F_1,\dots,F_n) = \frac{1}{Z} p({C}) \prod_{i=1}^n p(F_i | C)
\mathrm{classify}(f_1,\dots,f_n) = \mathrm{argmax}_c \ p(C=c) \prod_{i=1}^n p(F_i=f_i | C=c)


The settings can be modified using the settings browser.

Naive Bayes settings

  • (Layout)
  • Inputs: Number of input features.
  • Outputs: Number of output features(should be equal to the number of classes).

GUI details

The naive Bayes has a standard basic interface:

Naive Bayes training GUI
Image:Naive Bayes GUI short.jpg
Image:Naive Bayes GUI long.jpg

See also

This page was last modified 11:46, 28 January 2008.  This page has been accessed 6,662 times.  Disclaimers