@Kortuk Well I'm only living there until its sold, so technically I don't have to be able to afford it

@IvoFlipse :)

But at least I'll have my own office

Interesting site with free scientific journals: intechopen.com/subjects/computer-and-information-science

This might contain some interesting stuff for people: books.nips.cc (be sure to click through to each year)

Doing week 2 of stats

I haven't started yet

been rather busy searching literature and I tend to get lost :P

@IvoFlipse This one gets to Bayes rule, the foundation for all search engines!

w00t, I think I already got his explanation about that in ai-class

but still, I could use a basic explanation, hopefully it will make sure it sticks this time

@IvoFlipse :)

I am really annoyed with this question, he is not clear about his setup, I have done every other question without listening, minus one other he was not clear on, and this one is annoying me a lot, after listening no love.

and of course, after all of that, I did my math wrong, .4*.4 is not .24, no matter how much you want it to be

I am mad at myself

0.16 aint it?

What I 'hate' most is that I need to actually calculate stuff :P

And I take screenshot notes in Evernote, but I often have to time correctly to try and get shots without hands on it and make sure they didn't move the screen

@IvoFlipse :(

I should just stop taking such notes

@IvoFlipse I just hate the crazy number of problems we do, I would prefer one homework.

given that I hardly ever look back to them :P

@IvoFlipse Easy solution

@Kortuk Yeah, there are A LOT of questions in the current courses

Physics had that too

Which btw was a useful refresher on trig for me :P

@IvoFlipse stupid a lot

@IvoFlipse good :)

@IvoFlipse I use trig so often I think I am up to date.

I didn't have to use it for so long I just mess up all the conversions

plus if you've been taught Math in another language, its REALLY hard to memorize them

like there's no muscle memory left

@IvoFlipse haha, interesting point.

@IvoFlipse feel free to ask on these if you never need halp.

soscastoa vs Sohcahtoa

Even the names, hypothenus just doesn't ring a bell in my memory

I can just click through if I really get stuck

on this question he did not ask the question clearly and the page we answer on does not have the numbers you need for calculating the answer, you have to back up to find out.

During AI-class they had a ton of ambiguous questions and their reasoning was that it forced you to think it through and it didn't matter if you picked the wrong thing, as long as you tried

and i calculate in my head and it is not one of the answers, but he meant "close to"

But to hell, I want a straight A damnit

@IvoFlipse Ha

I should drop that habit too, if its costing me too much time

@IvoFlipse and to keep things "clearer" he is teaching bayes rule wrong first, I cant do this

Oh boy

I guess this just isn't for you

Coursera might be more your kind of thing

@IvoFlipse Still at the beginning, his teaching style does not match how I like these things taught at first

you're probably exposed to a whole lot more math to begin with anyway

I think the risk of writing a wrong equation once is never ever worth it, write it correctly and clarify you are ignoring a term to do a relative magnitude comparison since the base is the same(which is the case here)

Yeah, Khan Academy also does this

and now he is showing how to normalize!

if he makes a mistake, he'll correct it later

but the math is actually clear from looking at how the probabilities are written out!

which ruins my notes

Just make sure the material is correct already!

@IvoFlipse I do wish he would give the question written on the question page so that I could just jump through it.

this is only week 2, this is the basics

I am not sure I would attend the first 4 weeks of any statistics class I took in university.

I assume he is attempting to give an intuitive feel for someone whom does not like doing math or math is not a good place to found an idea.

and now I have a problem I dont know whether to solve for the real probability or the fake method he started with

Use what he has given you so far

@IvoFlipse but he taught the normalization step right after and it is the right way!

@Kortuk Then that's what he has given you ;)

@IvoFlipse it was pre normalization.

ohh, he writes it differently, i did not notice it before.

He writes P(c,neg) for the first step and P(c|neg) for the real probability!

Thank god, light at the end of the tunnel.

That's a gotcha, the second one is normalized right?

Its the breast cancer example I reckon?

@IvoFlipse just cancer in general, this is an example problem I argue with every professor about as oversimplification, they dont take into account the P(disease|doctor ordered the test)

the | means given

@Kortuk hehe

That depends on P(pay of doctor | doctor ordered the test)

P(coinflip is heads|given last flip was heads) should be P(coinflip is heads) in a fair flip, that implies there is no correlation between events.

@IvoFlipse Ha, yes, we have that issue with my job, doctors wont send patients to our clinic because they make more money letting them do the procedure in hours.

The business models in medicine / health care are so broken

@IvoFlipse yes, yes they are in many ways.

Worse, everybody has gotten used to making money off it that they'll protect it even more than bankers defend their broken business models

@IvoFlipse Here they preach free market, but in medicne it is treatment or death, you dont truly have an option to skip treatment many times.

@Kortuk Indeed, when on top of that the person who needs the help doesn't directly pay for the care he receives, that's a recipe for disaster

.021 does not count right if it was .0217

First it means that person doesn't mind if its more expensive, because the insurance company will pay anyway. There's no pavlov reaction that changes behavior when prices increase

Sure you'll bark when your insurance gets more expensive, but that's something you decide once a year, you don't know what you'll need next year and you have no clue how much your health the past year actually cost

And second, because you don't directly pay for it, you often don't care about quality as much as you should. So if the care wasn't as good as it could be or simply doesn't solve the problem as they promised, we don't call them out

@IvoFlipse I do, I feel there you always want the BEST because it costs the same so your sort for the showiest option.

@IvoFlipse We had a student project want to do something for a hospital and they found that it was worth making it cost 1000 dollars as comparedto 10 dollars if it made it look really cool.

because hospitals want stuff that looks cutting edge and expensive.

I'm mostly involved with people selling orthopedic shoes and orthotics, I find it disgraceful that they will sell you an orthotic when they know it won't help or sell a more expensive pair of shoes for their own profit. But what can I do about it

@Kortuk Hehe, make sure it looks like a scifi movie

@IvoFlipse Nothing :(

@IvoFlipse Ha.

Something tells me this measurement is going to shred my algorithm

@IvoFlipse Yes, Yes I think it will, is that a person?!

That's my grandma in her 80s walking over the plate, very very slowly

@IvoFlipse ha.

On a brighter note he is asking a few questions that if you have done a lot of statistics and found your intuition for it you can do without having to do math, I dont thing on purpose.

I have zero intuition for this stuff

especially when they ask the inverse of Bayes Law

@IvoFlipse No one does when they start. At least you recognize it. My last class the professor made a major point of saying you do not have intuition for this until you learn it, until I teach you axiomatic statistics your intuition will lead you wrong, do the math until you know the answer and the math always matches your intuition .

I blame my Math education for not making it more interesting

and my own disinterest in the topic :P

@IvoFlipse You have to like math to enjoy it :(

I enjoy what Math can do ;)

I find statistics unbelievably interesting, I found estimation theory to be one of the most interesting things I have ever done, just as I found DSP to be very very fun.

I'm loving computer vision, I just need more experience with actually getting it to work

and I seriously want to understand the statistics that go into fMRI

How on earth do I apply GLM to individual pixels?!?

@IvoFlipse Dude, FMRI gives me a nerder.

A raging nerder.

:)

> The crime of unlawfully killing a person by a nerd.

> A nerd boner or getting aroused by something that would usually only excite the most nerd-tastic

@IvoFlipse Ha, yes!

@IvoFlipse Not this.

@IvoFlipse I like to joke that my mind becomes rock hard and ready to do calculations.

animalsimulation.org/PDF/…

darn can't find a link to the article I'm looking for

Those are statistical comparisons between the everted and inverted footprints compared to a 'normal' one.

The top row uses a segmentation where it basically assigns one value to an entire anatomical region (average, max or whatever) whereas the bottom row compares each individual pixel. As you can see the inverted footprint has a significant increase in pressure when using the top method, whereas there are more pixels with a significant decrease when you look at the pixel level

@IvoFlipse hmm, interesting.

That's what I thought as well

@IvoFlipse Which works better is a mess.

Will be back online closer to morning your time.