Machine Learning and Cross-Validation

This lecture discusses the importance of cross-validation to assess models obtained via machine learning.
Book website: databookuw.com/
Steve Brunton's website: eigensteve.com
This video was produced at the University of Washington, and we acknowledge funding support from the Boeing Company

[@kadeeraziz]

You explain perfectly and eloquently with warmth. respects...

[@seongyongpark2347]

Thanks for your nice lecture, but I think you made one mistake when you explain cross-validation. You explained Monte Carlo cross-validation but you talked about k-fold cross-validation. In k-fold, we don't do random split k times but first split data into k divisions and perform one divison out cross-validation. If you can add comment about this in your video this would much better. Thanks your knit and facinating video once more.

[@2007beet]

Thank you so much. You're the best! I hope to finish most if not all of this series.

[@Eigensteve]

Enjoy!

[@spacewavematrix3009]

Thank you very much, Professor !! I'm learning a lot from your videos though being a slow learner, would like to watch all your videos as the concepts explained are with right examples and I can Imagine how well these can be correlated to real-world scenarios.

[@khaledsrrr]

keep them coming. I am loving ur teaching way

[@marxregis]

Professor, your videos are awesome. Could you do a video on sparsification factor lambda used in sequentially thresholded least squares algorithm?

[@geoffrygifari3377]

what do you think about these statements:
1. both interpolation and extrapolation are basically fill-in-the-blanks, with extrapolation the gap is in the future
2. the test data is what usually called a control variable in science.
thanks for the video!

[@sinamt2982]

Thank you so much for this good explanation!!

[@geoffrygifari3377]

is there a way to get the test data of the "highest quality"? (from your video i'm thinking if we have a continuous data across time, we set a certain percentage of those data as the test, sampled randomly throughout time)

[@geoffrygifari3377]

a complete newb here, are there places where you can get training data? (could be free, or a marketplace)

[@markghali7623]

If the physics/dynamics/system parameters could vary with respect to time or between the "Training" and "Testing" denominations set by the ML developer, is it possible for these changes to be treated as variables or inputs to better improve future predictions?

[@sutriadikurniawan4097]

how can the text can be like that? what apps that you use sir?. thank you sir

[@DiegoMachadodesigner]

Thank you so much!

[@TheBjjninja]

Why not mention the word "time series" or temporal structure or panel data.

[@supernova87a]

The biggest artificial intelligence here is how you write backwards in real time on that virtual board. How in the world...?