In leave-p-out validation, how many data points are used for testing?

In leave-p-out validation, p data points are set aside for testing in each iteration, while the remainder of the dataset is used for training the model. This technique allows for a comprehensive evaluation of the model's performance by systematically varying which specific data points are left out.

By utilizing p data points for testing, mean performance metrics can be calculated over multiple iterations, providing insights into the model's reliability and generalizability on unseen data. This method contrasts with other validation techniques where, for instance, only one point might be excluded (in leave-one-out scenarios) or where the entire dataset might be used for both training and testing without any specific p points designated for testing.