learningmachineslab / tutorials Goto Github PK

The tasks in the tutorial are appropriate for linear regression, but seem to assume a lot on the part of the participants. For example, even the first tasks just asks to develop 3 visualizations of the data to explain how things co-vary. Participants might need more scaffolding throughout to achieve the goals.

the data creation is currently residing in the data folder for each tutorial. this is sort of unwieldy and the data creation could be stored in its own folder in a single script to create the data for all of the different tutorials. this cuts down on repeated code and also mistakes in data creation.

Went through the regression tutorial and it appears to be focused on calculating quantities and making plots, but there are some important interpretation questions that we'd want them to reflect on that we should have answers to.

I'll start with the first task in which the solutions show 3 different visualizations that are not interpreted (i.e., answered the posed question). These other tasks are similarly missing interpretations from the solution.

all of the data thus far is just fake data generated from data generation libraries from sklearn. but i think its much more relevant to have data sets that are relevant to PER topics. This means like, a column shouldnt be feature3 but perhaps HSGPA or fci_prescore. then there can be discussions of the analysis of this data, like, i have no idea what it means that feature3 correlates to feature4, but if HSGPA correlates to fci_prescore we can think of some hypotheses as to why this is true.

subfolders can have readmes, this will be helpful to explain the function of notebooks in the folder, especially for users who are not in a workshop, course, etc. where they can immediately ask someone the purpose of a notebook. this should exist for every subfolder.

by convention requirements.txt/environment.yml are typically stored in the root directory. we should pick one. i think the environments file includes way too much we could probably just go with the requirements.txt

alternatively the reverse could happen since tutorials is much more aptly named.

the readme should be updated to be more public facing since it will serve as the tutorial repo for both the summer internship and conference workshops, etc.

with the replacement of all dthe data files with education simulated data, there should be no need to keep the old data files thus they should be removed

currently the .gitignore only works for the exploring data folder. it should work for all folders

for example, the logistic regression notebook in the classification folder can be removed.

the folder structure should be changed to match the exploring data folder structure. thus the R and python folders should be removed

example

topic/
├── data/
│   ├── data.csv
├── notebook1.ipynb