Nuggets of Data Gold - Computerphile

Data mining, why it's better than pure statistics. Professor Uwe Aickelin explains the basics of data mining.

This video was filmed and edited by Sean Riley.

[sight]

This guy just called Google a lookup table, that takes balls.

[arcooke]

That was incredible.  I watch every video posted on this channel and this one really struck me.  Can we get more of this guy?  Very interested in what he has to say.

[son-tchori]

One asks what is the value of one divided by three, then multiplied by three.
A mathematician answers "1".
A physicist answers (alternatively "0.99 ± 1%").
A statistician answers "¿ How much would you like it to be ?".

[NickHuntingtonKlein]

This guy has a very odd idea of what "statistics" is. No way of dealing with missing or unclean data? No way of uncovering nonlinear relationships? Absolutely not true. He's clearly knowledgeable about statistics but still seems to think these things, which makes me think he just has a weird definition for the term, which is nonetheless very misleading.

Interestingly, the causality discussion at the end of the video is currently where machine learning is weakest relative to standard statistical (and especially econometric) methods, so it's a very odd inclusion in a video with a subtitle about being better than pure statistics. Causality isn't completely unadressed in machine learning (see the work of Judea Pearl on causality in ML coming from the computer science side, or Susan Athey from the econometric side) but it's much less well established, and many of the engineering and CS people applying these methods in industry are largely clueless about these advances, or even the concept of causality outside an experimental setting.

[OwenPrescott]

There are some free MIT lectures on algorithms and AI worth watching.

[Friek]

This professors is one of the first germans whose english doesn't make me cringe. I even thought he was british for a few seconds.

[DanteD]

Data mining is just applied statistics... just without a hypothesis to start with. This is why it is much more prone to mistaking random patterns for causal ones.

[AstAMoore]

The true mystery is: What’s behind the gray rectangle mask over the computer display? We may never know . . .

[UberAlphaSirus]

Is there anymore info regarding the benefits of higher BP?

[hungbahuynh]

This video make remind me of tv series person of interest

[tsjoencinema]

Uwe Aickelin sound like just the man for James Harris "Jim" Simons company.

[MikaelLevoniemi]

Google and facebook do big data mining in much larger scale than professor here actually indicates. To get that data is easy, to piece it up is hard. That's one reason Facebooks aquisition of whatsapp is already profitable for facebook. They got the one piece they didn't have. Phone numbers. Billions of them.

[nosuchthing]

Amusing that a big data video appears a few days after some big data operations in the US are ruled illegal. Or alarming.

[ricardoamendoeira]

More from him and more data mining please!

[FYIitsnotme]

DRINKING GAME:
Drink every time he says Data
Drink every time he says Artificial Intelligence

[JustPingo]

What is hidden behind that square over the computer's screen?

[dixie_rekd]

professor Uwe has the most awesome accent im wondering where hes from.... anyone?

[Unplanted]

Dude, Paper doesn't grow on trees, you know?

[PeterWalkerHPc]

A common problem that analysts encounter is where managers ask for this and that mean or average. They come to us with their solution, not their problem. Thus, they often get what they asked for and not what they wanted;

[Haaknes]

This guy should teach statistics. I want more of these videos!