Digits: a potential little mistake in boundaries

Module 7. Evaluating model performance Choice of cross-validation
#1

Hi,
in the lecture:

writer_boundaries = [0, 130, 256, 386, 516, 646, 776, 915, 1029,
                     1157, 1287, 1415, 1545, 1667, 1797]

but:

data.shape

ouput:

(1797, 64)

that means we have 1797 rows for 64 columns
If we have 1797 rows that means index of rows go from 0 to 1796. I think the correct writer_boundaries should be:

writer_boundaries = [0, 130, 256, 386, 516, 646, 776, 915, 1029,
                     1157, 1287, 1415, 1545, 1667, 1796]

BTW that does not change the results after

#2

Not that the upper_bound is excluded in Python slicing so array[upper_bound] is not included in array[lower_bound:upper_bound].

I had to sit down to double-check but it seems like1797 is actually correct in this case.

#3

Hi
It s always good to sit down sometimes by day :grin:.
It s true that upper_bound is not included in array but here I was not sure boundaries were to design arrays as groups is built. If we changed the last number we will change groups in a way we have an addtional 0 at the end of groups and not a 13. My bad then.

I have an additional question about Digits:
If I m correct the data that were used in that lecture come from the test set of optdigits that has been extract from the Writer_independent set.
So you found writer dependencies in the writer-independent set?
Thank in advance for your help