search

accord-net / framework

Machine learning, computer vision, statistics and general scientific computing for .NET
http://accord-framework.net
GNU Lesser General Public License v2.1
4.48k stars 1.99k forks source link

SupportVectorMachine extrapolate solution #2206

Open mikesneider opened 4 years ago

mikesneider commented 4 years ago

What would you like to submit? (put an 'x' inside the bracket that applies)

Issue description

Hi, I am using a SupportVectorMachine model type Cauchy. The model was created with a minimum value to predict is zero, but running some scenarios the result of the prediction is a negative value. Question is, a model with SVM can extrapolate the prediction?

davidrapan commented 4 years ago

Hi, in theory it is more likely to happen with imbalanced dataset while introducing new unseen data to the model. Even when Cauchy doesn't depend on variance that much if I'm not mistaken. I would try some combination of kernels as it is the most common way and active learning. But of course it depends on what's your goal. And yeah.. I'm very new in this area and undereducated, so please take my thoughts very lightly.

D.