Do You Want To Learn Machine Learning? You Need To Be Good At Math

By Kimberly Cook |Email | Feb 21, 2019 | 3450 Views

Would people who are strong in math be good in machine learning? originally appeared on Quora: the place to gain and share knowledge, empowering people to learn from others and better understand the world.

Certainly having a strong background in mathematics will make it easier to understand machine learning at a conceptual level.

When someone introduces you to the inference function in logistic regression, you'll say, "Hey, that's just linear algebra!"

Then to optimize over that linear model they'll show you gradient descent, and again you'll note, "Well, that's just calculus!"
Support vector machines? Merely convex optimization.

Naive Bayes? Just probability theory.

But surely deep learning must be something new? Backpropagation?

Just more calculus. Literally. Not harder, just more (thank God for automatic differentiation).

Supposing you did well in those undergrad classes, machine learning should come easy enough at first. At least, until it comes time to actually build a model in the real world.

That's when all those software engineering skills that previously seemed so conspicuously absent from the equation start to factor in.

Automation, scalability, robustness, debugging, data sanitization, deployment, monitoring, reusability, version control.

At some point, machine learning becomes just another way of building software. And increasingly, the tools are becoming accessible enough that the theory and the math are becoming abstracted away behind frameworks that do all the heavy lifting for you (again, I can't stress it enough, that automatic differentiation saves the day).

So does being strong in math imply strength in machine learning?

Yes, without question, it makes one stronger in theoretical machine learning. Comprehending papers, implementing new approaches, understanding the frameworks under the hood.

No, increasingly, I would say that at the practical level machine learning is becoming less a research problem and more an engineering problem. When it comes to applying known techniques to real problems and deploying them into production, the skills that most enable one to be productive are technical skills.

That's not to say you should simply throw the theory out the window. It still helps, particularly when one needs to deviate from the standard playbook.
But in much the same way we no longer need to sort our own lists or balance our own binary search trees, we no longer need to build our own classifiers or calculate our own derivatives. And that trend is accelerating.

My advice? Whether you're strong in math or not, give machine learning a chance. But approach it from a different angle.

If the math seems tough, focus on the practical first, learn through analogies and by building something yourself.

But if the math comes easy, you're starting with a solid foundation.

This question originally appeared on Quora - the place to gain and share knowledge, empowering people to learn from others and better understand the world. 

Source: HOB