'Machine Learning Onramp' by Mathworks.

Just completed "Machine Learning Onramp" course by MathWorks.

Definitely one of the beautifully designed courses I've completed recently. It covers a lot of topics in-depth:

• Feature Extraction: The course covers in detail, the algorithms and programs used to extract readymade useful features from Data. The course concentrates on classifying Handwritten letters, and this module shows how images of these letters are rendered as matrices of values and used in MATLAB.

• Feature Engineering: Apart from the features that can be obtained directly from the matrices, engineering features that can be used to differentiate and classify letters are produced from the matrix values. Features like prominence points, speed, local minima and maxima, correlations and other features are produced and stored to help classify letters.

• Automating Feature Extraction: After these two steps, the course taught us to automate the above steps to aid easy reproducibility and also save time by automating these steps during preprocessing data itself.

• Training a model: The course uses ‘k-Nearest Neighbours (kNN)’ model and it’s variations to help classify the letters. Along with covering kNN in detail, the course also lets us explore other Classification Models using MATLAB’s Classification Learner tool in passing whilst concentrating on kNN for this course.

• Making Predictions, Investigating Misclassifications and improving the model: After fitting a kNN model, the course predicts some test values, and investigates the misclassifications, first by a confusion matrix and then in-depth by investigating the reasons behind the most common misclassifications by comparing their features using Parallel Coordinates approach.

The course focusses on Supervised Machine Learning and mainly on Classification. Overall the course content is beautifully designed in bite-sized modules. Looking forward to completing their ‘Deep Learning Onramp’ course to get a taste of Unsupervised Machine Learning.