Types of Machine Learning 1

This lecture gives an overview of the main categories of machine learning, including supervised, un-supervised, and semi-supervised techniques, depending on the availability of expert labels. We also discuss the different methods to handle discrete versus continuous labels.
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

[@hcordioli]

Excellent material for beginners and advanced ML practitioners. Clearly explained !

[@Eigensteve]

Glad you think so!

[@hannah_982]

Thank you for the video!
The way you write backward and the handwriting is better than mine

[@iftikharkhattak5914]

nice explanation and the way you write backwards is really fantastic. More power to you!

[@zns72]

This is just lightboard that is flipped in post. I do it all the time, but it does look cool.

[@zacharychristy8928]

Awesome content! You do a great job tackling these "next level" topics beyond just a surface description.

[@abdennourbouhounali2830]

Great explanation, thank you very much sir

[@mikiallen7733]

HI sir I liked your tutorials on fourier and wavelets yet would you kindly add a tutorial in the near future if possible on the quality of different ML methods / algos including RL and which one suits best time series data with the following empirical properties ; 1- non-stationarity 2- order matters 3- low signal to noise ratio so SD for example is a some multiple of the mean / central tendency measure , your input is highly appreciated

[@ai.simplified..]

As always great.

[@rishankdabra2034]

all the examples that you give to explain data science and ML concepts are from aerospace or mechanics domain except for cat dog example and those aerospace/mechanics domain examples becomes difficult to understand for a person not having that background so it would have been better if you gave simpler examples that don't require that or complex domain knowledge.

[@dippy9119]

In the example of "supervised learning with labelled data" given in this video, a labelled image of a cat goes in, the NN adjusts its weights and biases so a classification of "cat" comes out. But let's say I want to train a neural network so it can get good at playing tic tac toe against some opponent. When training this NN, the inputs are the values of each of the 9 squares of the game (each square is a cross, a nought, or empty). The output is the move to be made by the NN. Questions: Is this input data considered "labelled"? Is this considered "supervised learning"?

[@adamdudkiewicz6444]

dude i have been thinking about this comment for so long, and honestly i have no idea. i suppose that considering how simple the game is, given that you would be ready to spend a lot of time labelling the correct response to what is happening on the board, it could be possible. But then, i dont know how well would that work. For sure, reinforced learning would be a more obvious choice here, but supervised? hmm, maybe...

[@dippy9119]

@@adamdudkiewicz6444 using labelled data to "teach" a NN is called "supervised learning". This is not the only machine learning method. Another method, "reinforcement learning" could be used to teach a NN to play tic tac toe. This learning method doesn't use labelled data. Instead, the NN learns by receiving "reward signals" based on how accurate its output data is.

[@ai.simplified..]

@@adamdudkiewicz6444 maybe you mean semi-supervised

[@8Dbaybled8D]

when you realise he's writing backwards...

[@aayushparashar4143]

He is writing on glass

[@fackarov9412]

@@aayushparashar4143 naaah! he easily build a model in his brain to write in backwards

[@farmerfromwisconsin7256]

its a mirror, hes writing normal

[@8Dbaybled8D]

@@farmerfromwisconsin7256 if it's a mirror, why can't we see the camera?

[@chuntelee4371]

It's a glass for normal writing. Then we shall see the flip words. The reason we see the non-flip words is because the video was post-produced with flipping.