Machine Learning Tools or Concepts? where to start? Part 3

In the previous article, we started to talk about the topics and resources that will help you build models, and today we will add up more techniques in your algorithms’ toolbox while learning machine learning, the linear and logistic regressions.

Linear and logistic regressions: Those techniques, are considered the most widely used models in industry and academia, due to the nice properties of mathematical simplicity, computational efficiency and intuitive interpretation. you need to learn how the models are implemented using gradient descent. you will also need to cover the techniques when the assumptions of the model break.

The following points should be your objective while studying the linear and logistic regressions:

• Understand how objective functions of linear and logistic regression models are motivated.
• Understand the probabilistic approach of linear and logistic regression models: maximum likelihood estimation of objective functions.
• Understand regularization for linear and logistic regression: L1, L2, and mixed.
• Understand Stochastic Gradient Descent in depth, its mathematics, and how to implement it efficiently for Gigabyte in-memory datasets, both sparse and dense.
• Learn heuristics for making SGD work in practice, such as initialization and choosing learning rates.

Couple of a very good readings (and mathematically rigorous) for this topic are:

• Introduction to Statistical Learning (Chapters 3, 4.3) (a free PDF available online)
• Bishop, Pattern Recognition and Machine Learning (Chapters 3, and 4)

At this point of time, you are having the basic of ML, data processing and the mathematics concepts needs to move to advanced topics

After you are done with this. you need to start learning about deep learning concepts (Hopefully, I can publish something about this in a different story or posts, keep tuned!)

If you like to take courses, these are good courses and resources:

• Mathematics for Machine Learning Specialization (Coursera Course)
• Math for Machine Learning YouTube
• The 3Blue1Brown YouTube channel is really awesome one! And
• This book is also good (MATHEMATICS FOR MACHINE LEARNING)

Summary: in this part, we covered what you need to know in the context of linear and logistic regressions along with some good resources that help you understand the concepts in this part as well as the other parts of previous blogs. In the next article, we will talk about deep learning concepts and resources that helps you understand them on the right basis.