Thread by @svpino

If you are starting out with machine learning, these algorithms will give you the best bang for your money:

▫️ Decision Trees
▫️ Linear Regression
▫️ Logistic Regression
▫️ Random Forest
▫️ AdaBoost
▫️ Naive Bayes
▫️ KNN
▫️ Neural Networks
▫️ K-means
▫️ PCA

If you are looking to make things a little bit more practical, XGBoost will solve a lot of your problems.

I didn’t include it in the previous list because it’s a combination of Decision Trees with Bagging and Boosting, but it’s definitely one of algorithms that I use the most.

Information overload is a real problem. If you do a Google search, there are literally thousands of machine learning algorithms.

This list will keep you focused on the list that will give you the most benefits when you are starting.

https://twitter.com/EngFod/status/1375016779068674050?s=20

Unless you have a better strategy, tackle them in the same order they are listed here.

However, what will naturally happen, is that you'll see the need to follow your own path depending on your needs.

https://twitter.com/dillionburns1/status/1374992930625134592?s=20

I did skip it. Not because I don't think they are useful, but because I feel they are really complex.

Personally, I've always looked at SVMs as a cool theoretical algorithm, but I've never used them for anything.

Your mileage may vary.

https://twitter.com/__HenryMarx/status/1375009555348865029?s=20

Deep learning will only get you so far.

So far, everything I've built and deployed has something non-deep learning related.

A full system will likely be a combination of models working together.

https://twitter.com/fussyv7/status/1375073197570322434?s=20