If you only have a very small amount of training data, learning curves can be very unreliable and may actually raise an exception for very small sizes esp. if you have > 2 labels.
We should basically have the learning_curve() method raise an exception if it's called with less than 500 training examples.
If you only have a very small amount of training data, learning curves can be very unreliable and may actually raise an exception for very small sizes esp. if you have > 2 labels.
We should basically have the
learning_curve()method raise an exception if it's called with less than 500 training examples.