Roadmap to starting Machine Learning

It’s a very confusing world out there, with confused people giving out advices for free (much like the blog I’m writing XD). Every engineering student is learning cloud computing, data science, machine learning, blockchain or some other fancy sounding technology. Fortunately, or rather, unfortunately, we are in an ocean of resources, but just like the undrinkable water of the ocean, most of the resources available online does nothing to reduce our confusion. Not that the resources are devoid of knowledge, but finding quality study material in a systematic fashion seems almost next to impossible.

Drowning in the waves of random study resources, with the thirst for knowledge, one tends to take in whatever is found first. Like the YouTube video that promised you to teach you an entire programming language in under 10 minutes. (Seriously, who buys that?).

So after a year of wasting my tie trying to learn something worthwhile, my friend and I stumbled upon a video that we decided to try out. I mean who attends online college lectures, so why not? So here’s what you should start with if you are an ML enthusiast!

1. In-depth understanding of Statistics: Do you know there are two types of statistics ~ Descriptive and Inferential. Descriptive statistics deals with data to provide systematic descriptions of the sample, either in the form of tables or through graphs. Descriptive Statistics can further be broken down into Measures of variability and Measures of central tendencies. Inferential Statistics makes inferences and predictions on a given sample of data. If you were unaware about any of the above mentioned terms, you better learn statistics before you start with Machine Learning.
2. A thorough knowledge of Probability: I cannot stress how important the concepts of probability is in order to understand the various algorithms in ML. Start with types of probability (marginal, joint and conditional), move on to Baye’s theorem and then finally study Probability Density Function, Normal Distribution, Standard Deviation and curves and Central Limit Theorem.
3. Basics of Python (or R): Isn’t that obvious though? Without the knowledge of the basic syntax of a programming language, it would be impossible to understand, write and implement long codes of different algorithms and predictive models. Once you made through this jungle, you are free to dive right into the ML algorithms.
4. Theory: An extra advice from my side would be to first learn about Supervised, Unsupervised and Reinforcement Learning, and the application of the different algorithms inside it like Linear Regression is used to predict real values like cost of houses, total sales made, etc. Logistic Regression predicts discrete, binary values like yes or no, true or false, chances of winning, etc., and so on. Not only this, knowledge of concepts like confusion matrix and accuracy score will help in understanding the accuracy of the predicted results and the model fitted. Once you know the theory, you’ll know what exactly a piece of code is doing and why the particular model is being used.

Now you are all set to start your venture! Don’t be shy, dive right into the world of Machine Learning!

Signing off!