CVB Minos - Teach with LearnCLFFromMTS()? Parameters?

Programming Questions C-style API
1

I’m playing around with Minos and some of the pre-installed MTS that come with CVB (12.01).

When I try to use LearnCLFFromMTS() I don’t seem to get the same results compared to when I use the create classifier tool from the Teach Bench. When I load the classifier that has been created pragmatically in the Teach Bench, the classifier properties are different, some are even “0”, like “Indifference Radius” and so on. :expressionless:

This is probably due to the parameters of LearnCLFFromMTS()?! Looking at the documentation, some of the parameters are named Param1-6. What do they mean exactly? Can I change these?

2

Hi,

To get the same classifier from an MTS using the mentioned DLL function, it is essential to fill the TLearnControlStructure with the same values as the Teach Bench uses.

Please note that some of the parameters should not be changed at all. The Following values represent the default settings:

TLearnControlStructure

Param1 - Indifference Radius, default: 6
Param2 - Negatives Density, default: 1000
Param3 - minimum feature contrast, default: 8
Param4 - internal use only, don’t change, default: 15
Param5 - internal use only, don’t change, default: 2
Param6 - minimum feature count, default: 20

It is also important to keep in mind that the TLearnControlStructure, as a stuct in C# for instance, is not initialized with values. So, you have to set values for all parameters in any case.

1 Like
3

I see. You are right, I had not all parameters set for the control structure. I’ll try with all set to default first and start from there experimenting.