Intro to Machine Learning –– An Update

I recently wrote about how I started reading Introduction to Machine Learning with Python by Andreas C. Müller and Sarah Guido. So far, the guide has been quite helpful for learning the basics behind training a model––it offers digestible explanations, concise code, and references to every related source.

There are 4 key pieces of information that I have gathered thus far:

1. Many AI models are trained to make predictions for new data points based off of existing ones––often for categorization, advertisement, or other purposes. Specifically, a matter of classification is one that will sort a new data point into a group, whether thats yes/no or sunflower/poppy/orchid. A matter of regression will return a number based off a best-fit line.

2. In order to make the best of your data set, you must section it into two groups––training and testing. This can be a 50% 50% or a 25% 75% depending on your situation––just make sure to remember that a portion of your data needs to be sectioned for non-training purposes, otherwise you will run short on points that can actually determine whether your model works well or not.

3. To best label a new point, it is very important to have standardized, organized data sets. After all, the only reference for the new point is existing ones, which act as "neighbors" and thereby determinants.

4. It is crucial to choose a regression strategy that best suits your scenario––otherwise, you risk either overfitting or underfitting. Overfitting is a result of being overly complex––a model that is too reliant on the nuances of training data faces inaccuracy when new ones are introduced. Contrarily, a model that is too simple cannot capture patterns well-enough to make predictions.

I highly recommend this book. It is very beginner-friendly––Müller and Guido write with a clear, jargon-free tone and anyone can benefit from the integrated Python exercises.

Here is the link on Amazon for anyone interested: Link