• MLPY >= 2.2.0
• SciPy >= 0.7.2
• NumPy >= 1.3.0

You have to add the following lines to the beginning of the data set file to can be used with pyhnn3.

• Line 1: Type of problem, "cls" for classification problems, "rgs" for regression problems.
• Line 2: Number of attributes.
• Line 3: Character for missing values.

To define the type of each attribute you have to put:

• "int" for integer attributes.
• "real" for real attributes.
• "set" following of the possible values for nominal attributes.
• "ord" following of the possible values for ordinal attributes with "<" order.
• "bin" for binary attributes.
• "class" for class attributes.

For classification problems, class attributes have to be coded as 0-1 sequences.

Hepatitis data set:

cls
16
?
int
set 0 1
set 1 2 3 4
real
real
set 0 1
set 0 1 2
real
set 0 1
real
set 1 2 3
real
set 3 6 7
class
class
class

$ python pyhnn3.py -n {initial population} {dataset}