Module 1 - Week 2
Straight into details on Machine Learning this week and a subject close to my heart - trees !!!
In this instance 'trees' provide a useful analogy for data structures - from roots to leaves.
First we looked at 'Decision Trees' and then 'Random Forests'.
Again I was quite grateful for having some insight from developing web applications and using taxonomy structures, tagging and relational databases.
A lot of the presentation material had useful diagrams sourced from :-
Of course, being a tree enthusiast I liked the idea of Random Forests as the 'ensemble learning' process requires more trees.
In looking for imagery relating to Decision Trees and Randon Forests I found a blog post that explains a lot of what we covered in week 2 very well - Click here to view
The session then moved on to the type of formulas that gave me sleepless nights at school. This was an introduction to classification and regression. Click here for more info (without scary formulas)
Machine Learning Pipeline
Confusion Matrix
Practical Session 2
As we had got familiar with Colab last week our lecturer was keen that we did the code input this time after a brief run through of the notebook example.
Again, a familiarity with PHP functions, classes and libraries was useful in understanding the basic syntax and also use of code editors like Sublime and IDEs like PhpStorm helped me recognise the predictive text and auto-complete capabilities of Colab.
A quick search for Pandas and Scikit-Learn cheat sheets also helped me navigate through the exercises with relative ease and gave a basic understanding of how to apply Machine Learning to a dataset.
I've never looked at Python based systems before and it had me asking myself questions about why it is the preferred language for AI above PHP and whether you would even consider doing similar exercises in PHP.
The general consensus seemed to be that PHP is better for web development because it runs faster.
Of course, if I'm going to go any deeper and further with Python it did get me thinking more about CMS solutions like Wagtail on the Django framework.
