I love the English language. It is the means with which my family makes each other laugh. It is complicated, sometimes arbitrary in nature, I’ll give you that. But it a playground which spans an endless airfield full of hangers, with those huge fans that one practices skydiving over, or rooms full of old vases and lamps that you can obliterate using a baseball bat should you wish. Point is it’s vast in its application, English. Luckily, not only is a source of fun in my life, now it is used to do powerful things with computing machines.
Natural Language Processing
For this phase’s final project me and my partner, Brett, chose a multi-class NLP model solution. We used a Multi-Nomial Niave Bayes model to process around 9,000 tweets and determine their sentiment. A little over a year ago I wrote print(“Hello World”) for the first time. Now I’m Bayesian. It’s crazy. We used a pipeline with grid-search. I’m getting better at tuning hyper-parameters. I like the methodical testing of hunches and setting levels to extremes and also making tiny incremental changes. I wrote a function that allowed us to change only one hyper parameter at a time, based on sometimes no more than a feeling. The results of playing around with this function yielded an almost 10% increase in the accuracy of our model.
I don’t like them. Which means, based on past experience, I’ll spend a lot of time with them in the future. This is the way it is. sometimes you avoid mountain, and sometimes you climb it to make your legs stronger and your diaphragm more robust. And at the top there’s a great view of the sun sitting above the clouds to warm your body and shine on your eyes.
Neural Networks — They’re deep man
I remember searching YouTube last year and coming across some deep learning videos. I cached them away in a playlist, almost not daring to dream that I’d be able to build something like that. But in this phase that curtain was lifted and like the man in the painting, I saw beyond the horizon out into the universe. And it was good. And was the beginning. What else is there that I don’t dare to dream of achieving, that my future self finds himself doing and enjoys doing it? It makes you think. What are all the great possibilities, and can they really come true?
The Final Phase: A real life Neural Network
This phase’s NLP project leads to the final phase–phase 5, Capstone project time. And Guess what kind of model I’m building….