discretize data to learning naivebayes

[xchi]

Hi there,

I have discretized my columns of variables (except for the last column, which is already discrete since it is my class variable) from my dataset d.
for example, for the first two columns of d:

floatData = d.GetFloatData(0);
DSL_discretizer disc0 = floatData;
disc0.Discretize(DSL_discretizer::MethodType::UniformCount, 8, be[0]);

floatData = d.GetFloatData(1);
DSL_discretizer disc1(floatData);
disc1.Discretize(DSL_discretizer::MethodType::UniformCount, 8, be[1]);

I would now like to learn a naive bayes net from the discretized data. do i need to somehow write these disc0 and disc1 back into my dataset d?
because when I tried calling,

DSL_network result;
DSL_nb nb;
nb.classVariableId = "Type"; //Type is my class variable in dataset d.
if (nb.Learn(d,result)!=DSL_OKAY)

I get the following error at the last line...

Assertion failed: metadata[var].discrete, file d:\shooltz\dsl\repo\smilearn\dataset.h, line 186.

am I missing a step in between discretization and structural learning?

please advise, thanks!

[shooltz[BayesFusion]]

DSL_discretizer::Discretize returns the bin edges which you can manually apply to your continuous data in order to convert it to discrete.
It does not change the actual values in the dataset object.

While it's possible to use DSL_discretizer directly, it's much easier to call DSL_dataset::Discretize instead.

[xchi]

thank you so much. I used DSL_dataset::discretize, and it was so much easier!