Joon Young Kim, I agree that it should be $ \hat a_{ML} $ be $ \frac{|x_i|}{2} $.
Thus $ \hat a_{ML} = {max}_i\frac{|x_i|}{2} $
