ROC curves, or receiver operating characteristic curves, are one of the most common evaluation metrics for checking a classification model’s performance. Unfortunately, many data scientists often just end up seeing the ROC curves and then quoting an AUC (short for the area under the ROC curve) value without really understanding what the AUC value means and how they can use them more effectively.

Other times, they don’t understand the various problems that ROC curves solve and the multiple properties of AUC like threshold invariance and scale invariance, which necessarily means that the AUC metric doesn’t depend on the chosen threshold or the scale of probabilities. These properties make AUC pretty valuable for evaluating binary classifiers as it provides us with a way to compare them without caring about the classification threshold. …

TLDR; Do let me know in the comments or to my mail ID: rahul@mlwhiz.com, if you want to contribute articles on ML or Data Science. I would add you as an author.

Full Disclosure here: As part of the editing, I would try to provide you with comments and editing guidelines on your articles. I might also do some hyperlinking to other articles and add some links to my site and some courses as well. I might also add some blogs on my original site as well(mlwhiz.com).

A note on minimal formatting requirements for a post, apart from good content:

• Images: Add a main image at the top. All images in article must have source in their caption. In case you have created your own image, write in the caption: “Author Image”. …

I am a Mechanical engineer by education. And I started my career with a core job in the steel industry.

With those heavy steel enforced gumboots and that plastic helmet, venturing around big blast furnaces and rolling mills. Artificial safety measures, to say the least, as I knew that nothing would save me if something untoward happens. Maybe some running shoes would have helped. As for the helmet. I would just say that molten steel burns at 1370 degrees C.

As I realized based on my constant fear, that job was not for me, and so I made it my goal to move into the Analytics and Data Science space somewhere around in 2011. From that time, MOOCs have been my goto option for learning new things, and I ended up taking a lot of them. …

Machine learning is a growing field getting a lot of attention, but getting machine learning jobs is still very difficult. Landing an engineering role at a big company means knowing not just Data Science, but also things like programming and system design. More often than not, there’s a lot of research and learning involved to prepare for applying for a new position.

When I was preparing for my machine learning job interviews, I started preparing two months prior to the interviews. That’s when I really understood what I needed for the data science and machine learning positions I wanted. …

PYTHON SHORTS

Object-Oriented Programming or OOP can be a tough concept to understand for beginners. And that’s mainly because it is not really explained in the right way in a lot of places. Normally a lot of books start by explaining OOP by talking about the three big terms — Encapsulation, Inheritance and Polymorphism. But the time the book can explain these topics, anyone who is just starting would already feel lost.

So, I thought of making the concept a little easier for fellow programmers, Data Scientists and Pythonistas. The way I intend to do is by removing all the Jargon and going through some examples. I would start by explaining classes and objects. Then I would explain why classes are important in various situations and how they solve some fundamental problems. …

I love working with shell commands. They are fast and provide a ton of flexibility to do ad-hoc things. But the thing I like most about them — Oh, they look so cool.

Or so I thought until one day everyone around me was using shell commands. I just hated that development. The coolness had gone and working with that black screen was getting a little boring day by day. I had to fulfil my innate need to stand out. I needed that source of inspiration while shipping out that next code.

It’s not a cafe I am working at? Alas, people around me were getting to know that it was all pretty generic stuff I was working on. …

Office Hours

It was August last year and I was in the process of giving interviews. By that point in time, I was already interviewing for Google India and Amazon India for Machine Learning and Data Science roles respectively. And then my senior advised me to apply for a role in Facebook London.

And so I did. Contacted a recruiter on LinkedIn, who introduced me to another one and my process started after a few days for the role of Machine Learning Engineer.

Now Facebook has a pretty different process when it comes to hiring Machine learning engineers. They do coding rounds, system design, and machine learning design interviews to select future employees. Now as far as my experience as a data scientist was concerned I was pretty okay with the Machine learning design interviews but the other interviews still scared me. I had recently failed a Google interview for Machine Learning Software Engineer in the first round itself just because I was not prepared for Data Structure questions. …

Transformers have become the defacto standard for NLP tasks nowadays. They started being used in NLP but they are now being used in Computer Vision and sometimes to generate music as well. I am sure you would all have heard about the GPT3 Transformer or the jokes thereof.

But everything aside, they are still hard to understand as ever. In my last post, I talked in quite a detail about transformers and how they work on a basic level. I went through the encoder and decoder architecture and the whole data flow in those different pieces of the neural network.

But as I like to say we don’t really understand something before we implement it ourselves. …