Machine Learning

Coronavirus has seen many of us stuck indoors for potentially months so I’ve been taking advantage of the time trying out a course in Machine Learning.

I just finished reading Hannah Fry’s Hello World and realised that although machine learning affects a vast amount of my life, I don’t know a great deal about how it is actually implemented. Combined with my employer’s desire for me to take more courses, I enrolled in a couple with Udemy.

The specifically titled Machine Learning A-Z: Hands-on Python & R in Data Science paired with Deep Learning A-Z: Hands-on Artificial Neural Networks for a mere £28.98 seems reasonable so off we go.

The course seems logically structured and well laid out - today I spent about 3 hours on it covering data pre-processing.

One thing about Python that I find frustrating is that the language and it’s many libraries change frequently, often with little backwards compatibility. This was evident when using the Sci-kit Learn library after a number of changes including a deprecated method. In fairness this didn’t take me long to Google and fix but nonetheless it is a concern. The course leads are aware of this, having changed the downloads but not updated the course videos. Engineering can often see otherwise perfectly suitable solutions become difficult to maintain due to obsolescence and this is a similar problem in software.

Sci-kit learn

The course also has a couple of “draw the rest of the owl” moments. Starting multiple linear regression, there is a slide that says you need to know what a p-value is and links to two actually quite unhelpful sites. Fortunately Sal Khan to the rescue and I was up to speed. There’s another on one slide that mentions 6 occasions that regression is unsuitable but doesn’t explain any of them.

Previous post