I’m quite new to this topic, but I’ve set myself the task of understanding how to demonstrate that the average of prediction errors in the sample for a machine learning algorithm, which consistently estimates E[Y/X] (convergence in probability for each
X), converges in probability to zero.

I’m a bit lost and initially thought this problem would be simpler. Could anyone guide me on the steps to take, any additional assumptions needed, or recommend study resources to address this? Any help, suggestions or bibliographical references will be greatly appreciated.

Thank you in advance for your patience and collaboration!