e-classification

Assumption: You have a box which takes as input a classification problem and provides an e-optimal solution.

Classification problem = given examples of input x and binary output y, construct a function mapping x to y which is correct in the future.

e-regression

Assumption: You have a box which takes as input a regression problem and provides an e-optimal solution.

Regression problem = given examples of input x and real valued y, construct a function mapping x to y which is correct in the future.