Deep Thunk and Learning Machine Learning

Programming machine learning has been an interest of mine for decades. the last time I seriously looked into it, the technology just wasn’t there for the sorts of things I was interested in doing. Then I got busy doing other things.

Well, due to a conversation I had with a chip designing friend of mine, I thought it was time to give it another look. The available resources and programming tools for coding and using it in practical applications has grown incredibly fast in the past few years.

I’ll probably write at least a couple of articles on it, but as a “hello world” sort of thing to try and play with I decided to write up a pretty simple program that could use machine learning to train itself.

To keep things simple, I started with a nearby pile of data. Namely a bunch of articles and links to book chapters about Neural Nets and K Nearest Neighbors and the like which I intended on reading and studying.

I’m going to skip over the details of how it works for now, because I’m not sure what I’ll wind up doing with it. The gist of it is that you first put in the title, URL, and text for a bunch of articles. Then the program does a bit of hocus pocus with cleaning it up and counting words and a few other things, and spits out a list of suggested categories and what articles it thinks belongs in each category.

It sounds simple enough to do, but getting a computer to make decisions like that can be tricky. On the up side, when it works it can be particularly helpful in first pass sorts of sorting and filtering. Time will tell how things go with it, and how far I take it, but it has been a good experience so far, and I have a much better understanding of some of the theories involved, although I will confess some of the math is over my head.