Clear explaination of train and predict methods in ML library ?

asked 2017-01-31 19:19:13 -0500

Ihab Taher gravatar image

updated 2017-02-01 11:13:52 -0500

Can anyone refer me to a more clear ML Library documentation ? As I am confused now and can't get a real description of methods such as : Train and Predict. One more thing that I saw many times those train flags such as : Update weights and No input scale but I can't use them as the only definition I can see for train is this :

public  boolean train(Mat samples, int layout, Mat responses)
    {

        boolean retVal = train_0(nativeObj, samples.nativeObj, layout, responses.nativeObj);

        return retVal;
    }

So I want to know :

  1. Do I need to use train method in a for loop to update weights or I just use it once and it will perform what I need ?
  2. Does predict method returns the class label ?
  3. What is this definition of predict where it takes 3 parameters (Mat Test_Data , Mat Responses(output) , int Flags) and what is this Flags parameter ?
  4. How can I update weights during training or function train performs it ?
edit retag flag offensive close merge delete

Comments

1

hmm, there should be another overload for this, but also see this related question ... aaand this issue, too

berak gravatar imageberak ( 2017-01-31 19:37:05 -0500 )edit

Thank you. But unfortunately this isn't what I am looking for. I need a neat and clear explaination of these functions : Train , Predict and how to update weights.

Ihab Taher gravatar imageIhab Taher ( 2017-02-01 09:36:24 -0500 )edit
1

don't you like Machine Learning Overview ?

pklab gravatar imagepklab ( 2017-02-01 13:08:21 -0500 )edit

This is something good. But I can't find any explaination for train and predict methods.

Ihab Taher gravatar imageIhab Taher ( 2017-02-01 16:48:46 -0500 )edit