Listing 4.6: errors should be absolute values?

[mehaase]

The first line in the listing sums the error values, but since some errors are positive and some are negative, they have the effect of cancelling out and lowering the error measurement.

>>> avg_error = functools.reduce(lambda a,b: a+b, (trY2-Y_train))
10683.684611765864

I believe it should be:

>>> avg_error = functools.reduce(lambda a,b: a+b, np.abs(trY2-Y_train))
178465.4145040395

Or even more succinctly, reuse the error vector computed in an earlier step:

>>> avg_error = np.sum(error)
178465.41450403954