I have one confusion about the ratio which will affect the figure signifiacntly. When I try to multiply the ratio_train to 0.05, the train and test curve have similar trend. And the higher degree(complexity) may not influence the error nay more. So the train curve going up with the increasing degree is one particular case?
This is indeed correct. We chose to use a very small training ratio to exaggerate the variance of the output. If you keep the training ratio close to 0.5, there will be very little variance.
@sai_praneeth_reddy said:
This is indeed correct. We chose to use a very small training ratio to exaggerate the variance of the output. If you keep the training ratio close to 0.5, there will be very little variance.
The effect of ratio_train
I have one confusion about the ratio which will affect the figure signifiacntly. When I try to multiply the ratio_train to 0.05, the train and test curve have similar trend. And the higher degree(complexity) may not influence the error nay more. So the train curve going up with the increasing degree is one particular case?
Thank you in advance
Please add (a picture of) the results you obtain.
The first one is for default value (ratio_train = 0.005), the other one is for ratio_train equal to 0.05
This is indeed correct. We chose to use a very small training ratio to exaggerate the variance of the output. If you keep the training ratio close to 0.5, there will be very little variance.
2
Thank you very much
Add comment