GitLab

# Conflicts:
#	_episodes/01-introduction.md

There was still a value that referred to the number of features in the big dataset instead of the small one. Changed it.

extract content from episode 3 exercises

Among other minor edits, the spelling for one-hot encoding (https://en.wikipedia.org/wiki/One-hot) is now consistent across the document.

The installation command for versioned software, `conda install "tensorflow>=2.5" seaborn "scikit-learn>=0.22" pandas`, produces a conflict situation with the triplet conda 4.11.0, anaconda-navigator 2.1.1, python 3.7.11. An excerpt of the stderr message is: 

     UnsatisfiableError: The following specifications were found
     to be incompatible with the existing python installation in your environment:
     Specifications:
     - pandas -> python[version='>=3.10,<3.11.0a0']

Conversely, a plain command `conda install tensorflow seaborn scikit-learn pandas` installs successfully tensorflow-2.4.1, seaborn-0.11.2, scikit-learn-1.0.2, pandas-1.3.5, keras-preprocessing-1.1.2. All versions are newer than the minimum required by the course and than the version requirements that have been removed. 

The explicit version requirements are probably unnecessary and counter-productive. It may be more important to update conda in the first place, for example with `conda update --all`.

Co-authored-by: dsmits <djura.smits@gmail.com>

Co-authored-by: dsmits <djura.smits@gmail.com>

Co-authored-by: dsmits <djura.smits@gmail.com>

Co-authored-by: dsmits <djura.smits@gmail.com>

Use light dataset

Moving setting the optimizer to stage 5

Removed bbq dataset from CI

Generated new outputs so that the results match the results from using
small dataset

updating code-check-action branch

Added special installation instructions for windows