Collection of feedback

[stevehadd]

I have now used this material twice, all parts for the GMED/RMED training, and the first half for the CEDS 2022 conference. I will gather feedback here:

Feedback from participants:
• Generally, people who had a bit of background on ML found the session very useful, for me it was exactly what I needed to get started.
• The very beginner people in ML struggled a bit and suggested a few changes:
o the part on "history of ML at the Met Office" (which I actually quite liked but may be less useful for beginners) could be replaced by a brief overview of the different types of algorithms and maybe an example. (I know it's the focus of the 2nd session, but some people were a bit lost on this)
o For the Jupyter notebooks: we didn't have time to go through them ourselves much during the session, so potentially you could have demonstrated only one of the examples and then let us play with it.
o Some people hadn't had time to set up the conda environment so potentially sending it one week in advance would have been good (that said, again I appreciate you've done this for the first time and it'll be easier to send this in advance next time you run it - so this comment is probably not very useful!).

My own reflections:

• I landed up using a lot of the time, especially notebooks 1-3, talking with less practical time than intended. Probably need at extra hour for each session. So probably 4.5 hours per session (e.g. 10-12h30, 13h30-15h30)
• Need to find out about level of participants beforehand, to decide about how much background/introductory material to rpesent (e.g. deefinition of machine learning, what is supervised vs unsupervised). Might require something preparatory for participants beforehand to have consitent level - e.g. those new to ML watch DS CoP session first