In my previous post, I described a workflow that connects data with an app that visualizes the data.
There's one step left: automatically deploying the web app from GitHub to a server after every push to the repo's default branch.
In any project that shows data to users, be it documentation or a linguistic database, data must be separated from its representation.
This is almost an axiom. You have to keep your data in an easy-to-edit, non-proprietary format, while the tools with which you show the data to users can be as complex as needed. Tools tend to get outdated, and your data cannot depend on them.
I watched this video on automation by Lena Shakurova and I thought I could use a summary.
I have a Python library called tinybear that contains all simple tools I have used to work with files in different formats. It is published on pypi.
I was too lazy to write a blog post about publishing my project, but, in a nutshell, poetry makes these things rather easy.
Now, while moving in a direction of project management in IT, and having recently got a referral for a technical writer role, I decided, it was time to create at least some simple documentation for tinybear.
One of my goals recently has been to try and automate some of my PC routines. I didn't think it was possible at all, but then Lena Shakurova wrote/spoke about n8n.io, so I decided to try it.
This text was inspired by the section "Activation functions" in Google's Machine Learning Crash Course and a blog post mentioned in that section. I decided to explain the rationale of all this to myself.
Let's say we have a number of features and a binary classification task (e.g. whether a house will be sold, based on two features: its price and area).