Linear Regression + Decision Trees = Linear Trees

preview_player
Показать описание
Best of both worlds.

Рекомендации по теме
Комментарии
Автор

you made it so easy to understand. Now I feel I changed my mind about statquest. Thanks

MarwahTAli
Автор

Well, this is my evening sorted!!! Thank you very much

a
Автор

As always great job ritvik! Wow that is amazing, in less than 3 minutes you explained a pretty advance concept. Keep up the great work

ResilientFighter
Автор

Love this short format!

You can singlehandedly double the IQ of TikTok now too if you wanted.

cgabode
Автор

Great video. One quick suggestion, you might want to use different colors with drawing best-fit lines inside the same cluster.

BlackmetalSM
Автор

So we could essentially approximate any type of function with just linear regression

rudyorre
Автор

Wow, never thought possible the combination of lineal regression and decistion tree. It seems to be more powerful than piecewise linal regression methid. IS there anything like this in R?

TorTeCuenta
Автор

Nice. Any idea how to implement it in sklearn?

andresfelipehiguera
Автор

These videos are such incredible resources. You've taught me a ton--thank you!! I'm starting grad school and will share your link with all my classmates. For someone new to building machine learning models, do you recommend Python and Keras? If you've already made a video on getting started, I'd love the link!

danielm
Автор

Can this concept predict a higher value compared on what it was trained? Tree regression cannot, linear regression can. How about this linear trees?

RidingWithGerdas
Автор

Ive wondered about this before, but:
1) for this to be effective, you would want the tree to be fairly short, since running a linear reg over small amounts of data is not very effective.
2) How is this different from running some clustering algorithm and then a separate MLR on each cluster?

BaraPlaylist
Автор

Is this fitting a piece-wise linear function?

tianyuzhang
Автор

Really like these videos but do prefer the white board ones

Andres
Автор

Are there any major drawbacks with the 'Linear Tree' strategy? Given an experimental sample, is there any drawback to modeling n partitions using n functions vs modeling the sample with one function?

aZnPriDe
Автор

what is the local function you are talking about? Could you please mention some fun to determine that?

scar
Автор

Isn't the result just a linear model with categorizations of X as covariates?

danielezampieri
Автор

This tends to overfit on training data.

mattberg
Автор

Can I have your email if I got general questions regarding my research area?

MarwahTAli
Автор

Cool spin on a RT ... more parameters though

Wolf-hhrv
Автор

This is not a new idea. It's a predecessor to Logistic Model Trees (LMT) used for classification.

axlrose