Coursera Study Group

Phonon

11:34

@IvoFlipse Hey there

Ivo Flipse

Hellow :)

God I have to install Lyx (so I can write LaTex documents)

But it kept popping up dozens of dialogs for each package

Phonon

Are you running Windows?

Ivo Flipse

only now realized I could just tell it to install everything

Yeah

Phonon

I prefer MikTeX personally

Even though I'm primarily a MacOS person

Ivo Flipse

Well my client made the choice for me

Phonon

11:35

Ahhhhh

I see

Ivo Flipse

But honestly, I don't really care/mind yet

I've never used LaTeX before

Phonon

It

It's nice

I really like

it

Had to use in in grad school extensively

Ivo Flipse

They've asked me to help add references to the manual of their upcoming software

Since I've read nearly every article about pressure measurements up to ~2010/2011

Phonon

AHHHAHAHAHAHAHAA!!!!!

Just finished programming assignment 6 for algorithms

Ivo Flipse

Ah, very good

Phonon

11:44

with some optimization, I had a blazingly fast version of Papadimitriu's

Ivo Flipse

Hope you didn't get bitten by creating the graph the right way

Phonon

Some guy suggested preprocessing the input

Ivo Flipse

:O your skills > my skills

Phonon

No, it's literally this guy on the forum who made all the differenfce

Apparently there were only <100 variables that were usually relevant

If you preprocess the input, you'll be able to pull them out

Oh, I'm very satisfied = )

Ivo Flipse

Hehe, well done

I apparently had a simple working solution, but didn't create the graph the right way

Phonon

11:55

@IvoFlipse Yeah, I read a little bit about that algorithm

I didn't get completely why it was supposed to work, so I didn't spend much time on in

The local search solution was actually really easy to code

It's just picking random values in for loops and some bookkeeping

Ivo Flipse

Its really simple, but I had to add edges like so:

    G.add_edge(-x, y)
    G.add_edge(-y, x)

Phonon

right

Ivo Flipse

The bottom one, just means that the inverse is also true, which increases the likelihood of you finding an error in the logic

I didn't add that first -, which is REALLY odd

Because the minus (negation whatever) is supposed to be in the data itself :S

Phonon

Right

Ivo Flipse

But after that, you find the strongly connected components, like we did last year

Phonon

11:58

Yup

Ivo Flipse

Which just tells us with which vertices each vertex is connected

Phonon

I remembered how I solved that problem and decided not to go that route

Ivo Flipse

We store this is a hashtable/dictionary for fast look up

Hehe, I used a library :P

Phonon

I was solving it recursively (looking for leaders and what not), and kept getting stack overflows

Ivo Flipse

So I could just check for each vertex what its strongly connected component was

And see if the negation of each vertex was also present in the SCC

Phonon

11:59

I didn't feel like rewriting my code, so I rebooted into my Linux distro which gave me just enough stack space to run the program

lol

I figured I shouldn't test my luck with SCC because of that

Ivo Flipse

Yeah, that programming exercise was painful

I got bitten by mutable data structures :P

Phonon

lol

Ivo Flipse

Or rather that it wasn't clear whether Python was using a copy of the graph/explored set or the original

Phonon

Yeah, I forced myself to write everything in C for both parts of the class

except a few things that only made sense to do in Matlab

or another higher level language

Ivo Flipse

I wish I could do that :P

Phonon

12:02

I certainly couldn't when I started

That was my exercise platform

Now I'm much better at it

Ivo Flipse

12:19

Btw here's a course that seems to challenge the difficulty of PGM somewhat: alex.smola.org/teaching/berkeley2012/syllabus.html

I wish there was a MOOC based on that

Gaurav Jain

14:52

this is very helpful loria.fr/~rougier/teaching/matplotlib

Ivo Flipse

@GauravJain I've had to learn that stuff through painful trial and error

Gaurav Jain

@IvoFlipse ouch ... same and until I found this ...

Ivo Flipse

I can highly recommend IPython

Gaurav Jain

oh I agree with you

python = perl + R :)

Ivo Flipse

I normally used spyder, to have Python be like Matlab

Gaurav Jain

14:55

i am moving my entire code to python now ...

Ivo Flipse

Exactly, I'm watching the video's for Data Analysis, which is in R

Gaurav Jain

i am taking that course ... its really good course

Ivo Flipse

My eyes just bleed at some of the code :P

Gaurav Jain

hahahah

Ivo Flipse

Yeah, I thought it was just a rehash of the previous course

but luckily its proving to mix theory and practice somewhat

and my god does he have long video's

Gaurav Jain

14:56

i agree its ridiculously easy in python

Ivo Flipse

Anway, IPython rocks, because you can rerun parts of your code and keep images inline

Gaurav Jain

oh he has

i have to watch it 1.5x

can`t watch 2.0 x and 1.0 is too slow

Ivo Flipse

I'm typing along with his video's, so I'm really not even listening

I mean, the stuff is pretty straightforward

Gaurav Jain

ipython ... never heard if it

Ivo Flipse

had it been a blog post or a book, I wouldn't need the video's

Gaurav Jain

14:57

of*

Ivo Flipse

ipython.org

Really useful when demonstrating some code, you can easily share the code and have people run it line by line or block by block

I've seen everybody use it in the PyCon video's

Gaurav Jain

yeah ... this looks cool

Ivo Flipse

Basically you create a server (either locally or a real one), then you run kernels on it (which is just your script)

But it clearly scales to much more complicated stuff

Gaurav Jain

well most of our code is still in perl and R and its a legacy so I can not change much ... but i have move most of my code to python

Ivo Flipse

It has a lot of tools built-in, so its easy to use for scientific programming

Gaurav Jain

14:59

i`ll read about it more and use it soon

Ivo Flipse

@GauravJain You could write regression tests

Gaurav Jain

thats really cool

so howz life ...

Ivo Flipse

Life is good, though the coming month I won't be programming as much as I would like to, but such is life

And yours?

Hmm I didn't know about GridSpec

Any new Nature publications? ;)

Gaurav Jain

tough ... we are moving to a new building next month .. and I have many projects that I want to finish soon ... publish some paper as a lot of people are working on and I don`t want to loose our edge on that

hahaha not so far ... but lets see

yeah that GridSpec is pretty cool

you were all along right ... python is good :D

Ivo Flipse

Oh and the origin for imshow is really important

I got bitten by it, because all my measurements were displayed upside down

Because I didn't know what left or right was, I didn't know that beforehand

@GauravJain I'm very happy I picked Python myself

Gaurav Jain

15:03

Bill goes to a library and asks for a book ” Psycho, The Rapist”.
The librarian searched for hours… comes back slaps Bill
She says ” Idiot! The book is called Psychotherapist”

Ivo Flipse

Though apparently my client thinks that makes me 'not a real programmer'

Gaurav Jain

@IvoFlipse well it was a wise decision :)

Ivo Flipse

Mostly driven by the fact that Matlab costed $$$

But compared to Matlab, Python is a breath of fresh air

Even though Matlab has excellent docs

Gaurav Jain

@you can`t get stuck in 1970`s programming scenario ... things have changed and came really far ... need to use all the new programming tools to make life easier and faster throughput

Ivo Flipse

I'm not really bothered by it, I can do plenty of useful stuff with my 'fake programming language'

Gaurav Jain

15:05

I agree ... i havn`t use matlab that much ... but still python does the same job ... and its FREE ... i don`t have money to buy matlab ... so i used octave .. but still

hehehe python: the real fake programming language

Ivo Flipse

Fake Programmers

The room for elite Super Users who realize they know nothing a...

<- proud Fake Programmer :P

4

A: How can I speed up an animation?

I found Joe Kington's answer that mentioned using Glumpy instead. At first I couldn't get it to work on my own data, but with some help on chat we managed to figure out how to adapt one of the Matplotlib examples that come with Glumpy to work on my data. import numpy, glumpy from glumpy.pylab im...

I just need to figure out how to fix this thing (again)

I need to embed that Window into a wxPython window, but the backends don't seem to match :\

And I don't really like PyQt that much

Gaurav Jain

this is cool

Ivo Flipse

I have it working in my PyQt app, but I'd prefer to be able to use it with any backend

@GauravJain Why the hurry to finish them soon btw? Worried someone else will publish something similar?

Gaurav Jain

20:55

yeah something like that :)

Fake Programmers

Transcript for

Coursera Study Group