Support vector machine

So any tree-based algorithm, decision trees, boosting, gradient boosting, edge boost, anything regarding neural networks, logistic regression, linear regression.

We had SVMs.

Those are not as popular anymore, but I think they're beautiful.

um and uh and then we had a fair amount of something i find very important which is kind of how to evaluate the models right like it's not just train the model but like is it doing well do i just add more layers and and that's it or do i need to check other things like is it is it performing well in my data set outside of my data set uh metrics so we did um cover a lot of metrics, overfitting, underfitting, all the keywords for like a machine learning interview, basically.

Want to understand, uh, because you have a front row seat on this, is that, uh, general machine learning was done, uh, has been done I think over, I don't know, close to two decades now.

I mean, your things like SVM, SVM and your, uh, your classification networks, your computer vision-

Yeah

... things like that.

Option next is called support vector machine.

They find the best boundary that separate one class of data from another.

It mean if the question talking about identify image, image classification based on decision boundary, answer is basically SVM.

Next is called K-nearest neighbor.

It classify based on a similar nearby data.

And I would argue that what we denominate as reasoning, the reasoning capabilities of the models have been another discontinuity.

So you have had like bang, bang, two massive shifts in the technology on top of a lot of boring and very effective technology behind it, which, you know, we've talked about this many times, which could be boosting decision trees, SVMs.

You still have technologies out there that continue to be relevant and very useful to enterprises.

If you look at industry, for instance, or manufacturing, there is still on the shop floor a lot of non-neural network technology being used.