iacobo

Added logic to automatically update links for all episodes, by using video id of top search result for episode title on YouTube. Also tidied up the pipeline using a DataFrame.

Cannot fetch 'MINST original' dataset (or any dataset using fetch_mldata) - website down permanently

2

As seen [here](https://github.com/scikit-learn/scikit-learn/issues/8588), `sklearn.datasets.fetch_mldata` has been deprecated as of version 0.20 and will be removed from sklearn in version 0.22 since **the source website mldata.org went down as of March...

If one sequence has e.g. only CGT the old code will inconsistently assign integers to each letter if other sequences have a different set of letters (i.e. ACG or ACGT)....