Coding gaussian process regressors FROM SCRATCH in python

In this video we will implement a Gaussian process regressor with squared exponential kernel in Python using numpy only and code several interactive plots to visualize it. Feel free to adapt the code and get your own hands on GPRs!
Easy introduction to GPR (uncertainty models): • Easy introduction to g...
0:00 Intro
0:43 Preliminaries
1:35 Implement squared exponential kernel
3:48 Implement GPR
12:48 Plot GPR
15:05 Draw random functions from GPR
17:08 Add points iteratively
18:07 Change parameters
LinkedIn: / nicolai-palm-97160b219
-----------------------------------------------
Data Science to go: paretos.com
Complete Notebook: gitlab.com/youtube-optimizati...

[@OliverJanShD]

This is amazing! Thank you for providing this approach, it really helped me understand GPR a lot better

[@reallyanotheruser7290]

Out of what feels like two dozen tutorials and explanations i found this is actually what made me understand it

[@youngzproduction7498]

This is a solid gold for me. I like learning anything in a visual way which I can interact with it. Thanks for your effort.

[@ross302ci]

Extremely helpful for understanding GPRs, thank you!

[@umutkorkut8555]

Very cool and easily digestible content, loved it!

[@33gbm]

Excellent material you provided here; I just came back to the video to congratulate your efforts on the content hahaha Thank you, man!

[@brendarang7052]

Nice one! Thank you.

[@TaylorSparks]

underrated video! Thanks for making this great content. This helped me quite a bit as I prepared a lecture on this topic for my materials science students.

[@nicolaipalm7563]

Thanks! Glad I could help you 🤓

[@muhammadrayyan4389]

You are amazing!!!....thanks for helping me in studying for my Green Light meeting which is due in less than 2 days!!!!..this video gave me a great confidence!!!!...once again thank you very much!!!!!

[@farhadazadi]

This was fantastic

[@amalroymurali3457]

Excellent stuff. Thanks!

[@pouyaaghaeipour8336]

Can we have access to the notebook file?

[@amothe83]

Excellent , the best video on gaussian process regressors

[@kaushikgupta1410]

thanks for this amazing explanation

[@swisscheese9590]

Great video. Would you do a follow-up on hyperparameter optimization using marginal log-likelihood in the loss function?
Also, a visualization example using multi-input GPs would be interesting as well. Or multi-output GPs.

[@satadrudas3675]

This is so underrated. Good job anyway

[@eva__4380]

This was really helpfull for me in understanding GP thankyou so much for your efforts

[@icoop99]

Very nice video - thank you very much :D

[@komuna5984]

Absolutely Mindblowing Work! Keep it up. May Allah bless you. 🙂

[@fatismajli2490]

Nice Video!

[@azd.zayoud]

Well done!
# Writing comments would be helpful for beginners
if it is put in a context of solving a problem/examples :
it will be more useful.
Thanks!

[@MrSchwede91]

Kommt die Fortsetzung noch? Bisher alles sehr gut beschrieben...

[@jaimesastre6393]

Amazing 👌🙏👌
Access to the notebook would be great 🙏🙏🙏

[@nicolaipalm7563]

thanks! 😀
link to the notebook is in the description

[@yeshuip]

hello,
how to feed sequence of input data to train sequences of outputs

[@nicolaipalm7563]

With this framework you can feed multidimensional input to the GPR. In order to obtain multi dimensional output you simply train a GPR for each component of the output vector. :)

[@yeshuip]

@@nicolaipalm7563 thanks for the reply.
but my question was is it possible to feed N*d matrix as input and N*2 aa output. where N represents the input sequence and d represents dimension of features and N as output sequence number and 2 as number of output features

[@IvanStar96]

Is sigma 0 or 1 in this example?
The title of the graph says it is 0, but doesn't the code say it equals 1?

[@nicolaipalm7563]

Yes that is correct. 😀

[@junaidlatif2881]

Can we use your method on our data?

[@paretos-com]

sure! What kind of data is it?