Add (Quick) How-To Examples

[seanlaw]

Recently, I received some feedback:

1. There is too much friction to "get started". With limited time, the tutorials are too verbose for users who are looking for the one line of code to use the package (i.e., stump(data, m)).
2. It isn't always clear what the shape of the input should be in
3. P_Point_Point = stumpy.stump(point_df["Centroid Location"], m)[:, 0].astype(float) is hard to read
4. stumpy.stump is somewhat long. Consider st.stump when importing
5. how quickly does the documentation allow me to implement everything on my data as someone that has spent 2 hours learning about the methodology
6. The tutorial titles are named by someone that has expert knowledge and may be off putting for beginners. Instead, the title (or description) should quickly tell you what the tutorial does

as a data scientist i have data and a problem, and am looking for a solution tell me what solutions you can offer me is my frame of reference

Ultimately, from the user's perspective, STUMPY is package that they need to figure out how to use in as little time as possible. Even if one package is "better" under the hood, they'd probably accidentally end up using it, if I can get it to work on my data faster.

this is probably the document that got me to spend more time with the matrix profile https://www.cs.ucr.edu/~eamonn/100_Time_Series_Data_Mining_Questions__with_Answers.pdf selling me solutions, not a new shiny methodology

I guess to summarize:

1. get user motivated/excited (what can the thing solve for user)
2. get it in users hands and working
3. now they're more invested and more willing to get into the details

I think it would help to have "Quick Start" guide with simple examples of how to use STUMPY that avoids lengthy explanations. Perhaps, we can create super-short versions of our tutorials and convert them to quick start examples.

[namankaushik9876]

Hi, I want to work on this issue. Give me chance

[seanlaw]

@namankaushik9876 Thank you for your interest. Perhaps, the first thing you can do is help us identify other docs (from other packages) that could serve as a reasonable example to follow and that we could mimic the style of.

As notes above, I think these will be short (maybe 2 paragraphs?) and the code should focus on the STUMPY API and have east-to-understand titles so that experienced data scientists (who know what they want to accomplish) can quickly use the minimum-reproducible code snippet. This should not include code for plotting figures and, instead, we simply embed the necessary figure. But first, we need to find short examples to compare between.

How does that sound?

[namankaushik9876]

sure sir

[seanlaw]

Here are some crowdsourced references that may be useful:

1. Dividing up documentation: Tutorials and How-to Guides
2. Scikit Learn Tutorials vs Examples - I like the cards a lot (ignore the "Releases")

[NimaSarajpoor]

this is probably the document that got me to spend more time with the matrix profile https://www.cs.ucr.edu/~eamonn/100_Time_Series_Data_Mining_Questions__with_Answers.pdf selling me solutions, not a new shiny methodology

Related issue: https://github.com/TDAmeritrade/stumpy/issues/107