In the function `LinearLearner` in `learning.py`, `learning_rate` is not used in the gradient update rule.