Supervised learning vs unsupervised learning.

[Human Epic Jolt]

Have at it.

How do you replicate? How fast?

Where is it going?

[xjoex]

For fun I teach computer science grad school for a state university. 80% of the class is in person, they film it and some distance students watch it online.
It is a totally different experience for the two types of students. Some do really well in person, asking questions, and they need the rigor of class to actually stay on top of it. Some don't need that at all and can do fine.

But whatever fits your learning style. I can learn almost just as well from a book as I can from a class.


-Joe

[Human Epic Jolt]

80% of the class is in person, they film it and some distance students watch it online.
It is a totally different experience for the two types of students. Some do really well in person, asking questions, and they need the rigor of class to actually stay on top of it. Some don't need that at all and can do fine.

Sounds a bit complicated.

Let's stick to machines.

[SpeedyChix]

Sounds a bit complicated.

Let's stick to machines.

Working with tools, working on machines, etc.: Learn the quickest by observing someone so I suppose that would be considered supervised learning? Though generally while watching it was just from hanging around, not specifically watching to learn. The "how to" just sticks. I attribute my mech. and troubleshooting ability to having learned that way from a very early age. I'm the "wrench" in the house when it comes to building and maintaining bikes. Computer hardware and LAN work, all picked up by watching and then going on from there. Reading "how to" do something is a much slower process for me when learning a new skill than actually doing it.

[mjbabcock]

There's a lot of stuff done on cognitive apprenticeship and communities of practice. Lave&Wenger and Brown, Collins, & Duguid have some cool reads on learning. In the end, supervised by someone who scaffolds the experience is usually better

[treadonme]

You'll find out Dec. 21

[Human Epic Jolt]

cheers. scaffolding and lan = good.

im thinking when a machine can pick it's own control and how it got there vs how long it takes to find out.

[Human Epic Jolt]

vs

what you know.

[Human Epic Jolt]

i think im talking hardware speeds.

[spopepro]

Well, I guess it depends on the quality of the training set, and the specificity of the hypothesis set. Also how stochastic the data set is. Chances are, no machine is going to learn much unsupervised, unless there's something interesting going on in neural networks research that I don't know about (I'm just a pure math guy...)

We are talking AI, right?

[xjoex]

My bad, I wasn't thinking machine learning. I did do some work with weka but not my cup of tea.

-Joe