Why linear regression is so important #shorts

-- -- (Links on this page my give me a small commission from purchases made - thank you for the support!)

Here's my favourite resources:

Best Courses for Analytics:
---------------------------------------------------------------------------------------------------------

Best Courses for Programming:
---------------------------------------------------------------------------------------------------------

Best Courses for Machine Learning:
---------------------------------------------------------------------------------------------------------

Best Courses for Statistics:
---------------------------------------------------------------------------------------------------------

Best Courses for Big Data:
---------------------------------------------------------------------------------------------------------

More Courses:
---------------------------------------------------------------------------------------------------------

[jmgpqcx]

To extend it a bit, a shallow neural net (1 hidden layer) is simply a piecewise linear function with some joints in the 1D input space.

[Cin]

As a European this is massively confusing that you guys dont know 7th grade math…

[edsonwinnerify]

If there is a reason for me to keep learning programming is youtubers like you. Thanks!

[Yebjic]

AI is just lazy optimization. Change my mind.

[tharakaratnayake]

Actually, a little bit of difference is there other than more numbers.

Unlike Linear regression, neural networks can capture non-linear patterns using their activation functions.

[jayleo]

I think understanding what is means to minimize entropy was what helped me the most with machine learning.

[bobinwobbin]

“With only more variables” well except that the activation functions are non linear and out of a purely pragmatic position render learning linear regression for the sake of ML pretty useless

[handmadesoap]

And that's why we need extremely good gpu for commercial models.. we need better way for training

[perisaizidanehanapi]

linear regression assumptions makes it less appealing, so people usually just overstep it and carry on with their model.

[mecha_au]

I would really recommend visiting your local barber and asking for suggestions on how to better style your facial hair. Sorry bro.

[ericb]

I've never heard beta_0 + beta_1 * x = y. I'm used to:

f(x) = a*x + b
or
y = a * x + b

[brendanlydon]

Recent subscriber. Great videos. I was wondering if you could do some videos on direct comparisons between support vector machines and neural networks. In my experience, using the RBF with strong regularization parameters is much better than the neural nets I train. I deal mostly with non linear relationships with 3 class imbalanced datasets. Ultimately, I’ve yet to realize the hype around neural networks for my problems and it’s bothersome. SVMs, however, have pleasantly been everything I thought neural networks would be. What are the best activation functions to use for nonlinear, multi-class neural networks? I already saw your fraud detection video for imbalanced datasets comparing 5 major models.

[pedroantunes]

Linear regression is not only a LINE OMG. It's ant regression that is linear in the pararameters, can be a 120-th order polynonia

[roeeorland]

Except in NNs we add non-linearities (even for simple multi-perceptron networks).
Plus later we have dropouts and other fun stuff

[kennedymwangi]

I just learnt so much in such a short time.

[complexity]

This man packs alot of info in his shorts.

[CC-.]

Doing a neural network is more complex

Just like how it's easy to manage small amount of items than large amount in suppose a database facility