>>13829947I work in ML, and you're conflating it a little bit too close. Classic statistics and ML are not super-interchangeable; in a huge generalization, statistics is about reducing the number of parameters and explaining feature impacts on results, while ML is about predictions- more parameters the better.
>"Machine learning" "deep learning" "recurrent models" etc are all just regressionis a little absurd; you could probably twist definitions enough to consider it half-true, but I wouldn't put any real philosophical weight behind it.
Obsessing over linear regression is a little trite; its like telling someone who wants to build a bookshelf to go into the woods and study types of trees and the metallurgy of hammers before buying some lumber. Certainly you want to know the basics, and I absolutely agree if you were highly focused on the statistics side of things (explanatory vs predictive), but I would suggest something different.
>Also, people will recommend you Python, I recommend R. There's a reason statisticians use it over python. Follow what the statisticians doI think you put too much weight in statisticians. Again, they are solving different problems. I worked 100% with R for my PhD, and now I almost exclusively use python for my job (aside from graphing; ggplot is too good). It really depends on what environment you are in and what you are trying to do. I can't do in R what I do in python because certain packages just really aren't there. But R is definitely one of my more favorite languages I've ever used.
>>13831018If you want to get deep into ML, go to kaggle and look at all their example notebooks. Like drinking water from a firehose, but you'll notice very little linear regression in any of it when it comes to solving real world problems.
>>13829832brings up a good point that is often thrown at the codemonkeys masturbating over their new algorithms, but don't swing the entire opposite direction and think most modern ML is useless.
