Journey Through the World of Machine Learning Algorithms

A beginners guide to learning machine learning algorithms

A major chunk of Machine Learning is Algorithms.

You might have mastered cleaning and tidying up the datasets and filtering out the features. What’s next ? Well, moving forward with the modelling, you will need to make a choice between different “possible” algorithms your dataset can be modelled on. More often than not, you will have to combine several algorithms to create your models in order to get the desired results.

In this short article, here are my five approaches to learning machine learning algorithms, as beginners.

1. Compile a List of Machine Learning Algorithms: To begin, create a list of machine learning algorithms, categorizing each one based on its general category or type. This exercise will help you become familiar with the various algorithms available, and as you gain experience, these lists will serve as references for your projects. You can find algorithm lists in books, articles, or websites like Wikipedia.
2. Apply Machine Learning Algorithms: The best way to understand machine learning algorithms is to apply them to real datasets. Practicing applied machine learning will help you bridge the gap between theory and action. You can work on problems that interest you, explore competition datasets, or use classical machine learning datasets. Utilize machine learning platforms like Weka, R, or scikit-learn to access and experiment with various algorithms. By doing so, you’ll develop an intuition for different algorithm types and learn about their preconditions and parameter effects.
3. Study and Describe Machine Learning Algorithms: Before diving into the details of an algorithm, explore what is already known about it. Research the primary sources where the algorithm was first described, consult authoritative interpretations in textbooks or review papers, and explore conference papers and competition results. While researching, build a description of the algorithm using a template, which can include references, pseudocode, best practices, and usage heuristics. This process will allow you to create your own mini-encyclopedia of algorithm descriptions for future reference.
4. Implement Machine Learning Algorithms: Implementing machine learning algorithms from scratch is a valuable way to gain a deeper understanding of their inner workings. When you code an algorithm yourself, you’ll make many micro-decisions, some of which may not be exposed through configuration parameters. This process will help you customize the algorithm and gain insights into the mathematical descriptions and extensions. While your initial implementation may be less scalable and robust compared to production-grade versions, tutorials and open-source implementations can aid you through the challenging parts.
5. Experiment with Machine Learning Algorithms: Treat machine learning algorithms like complex systems and study them as a scientist would. Conduct experiments by controlling variables, using standardized datasets with known characteristics, and analyzing the cause-and-effect relationships of algorithm parameters on results. This understanding will enable you to configure algorithms more effectively for future problems and adapt them to different domains. Keep in mind that many machine learning algorithms are stochastic in nature and require empirical investigation and probabilistic descriptions for better comprehension.