[FoRK] Raspberry Pi + Qt5 + OpenGL = awesome

Stephen D. Williams sdw at lig.net
Fri Apr 12 01:05:48 PDT 2013


On 4/11/13 11:10 PM, Joseph S. Barrera III wrote:
> On 4/11/2013 10:01 PM, Stephen D. Williams wrote:
>
> > Since a lot of the interesting machine learning is done with Matlab/Octave now, I'm trying to imagine that much Lisp doing 
> similar things...
> > I know about the early stuff, Mycin etc.  But there is a big gray goo period that I'm glad I skipped over, or rather skimmed 
> through, jumping to the recent good parts.
>
> This is a big area of ignorance for me.
>
> Most of what I know of Matlab is:
>
> 1. We keep trying to get the scientists at SLAC/LCLS to use (free) Python/scipy/matplotlib/etc... instead of Matlab but they won't.

My sense was that I'd want to see a concise equivalence to each key feature of Matlab / Octave.  The concise syntax becomes an 
algebra that you think in, so using Python instead is a bit disruptive; perhaps OK if it were easy to see equivalence.

The most surprising thing about Matlab was a kind of free, implicit polymorphism where you could create functions that did complex 
operations where the function would work on a wide range of forms of input.  I should dig up the exact example from a machine 
learning solution, but it was something that would have taken a lot of alternative code in Python or Java most likely.

Matlab is a pain:
https://abandonmatlab.wordpress.com/2009/07/

Why isn't there an .m syntax parser for Python?
I wonder how well this works?
https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2644618/
http://ompc.juricap.com/
Rats:
> The stage of the development still doesn't allow you to use it as a replacement for MATLAB(R). 
Perhaps this works:
http://chiselapp.com/user/victorlei/repository/smop-dev/wiki?name=SMOP+--+Matlab+to+Python+compiler

SLAC should have a Summer of Code project for this...

Or merge Octave and Python somehow.  Perhaps:
https://pypi.python.org/pypi/oct2py

Seems like SciPy only loads .m data:
http://docs.scipy.org/doc/scipy/reference/generated/scipy.io.loadmat.html

Recently, I wanted to do some Python scripting on MacOSX and was very surprised to find out how disjoint and annoying it was to get 
a newer version of Python installed and be able to easily install Python libraries that would be available for scripts.  I tried 
several published mechanisms and finally gave up, just setting the python path to the directories of the libraries.  Seems like it 
is only solved for Linux plus Windows with a commercial distribution. Disappointing.

> 2. We have tried getting the scientists to use Octave instead of Matlab but like no way.

Probably: There's some UI and other capabilities that aren't there on Octave.  And maybe not trusting that it is as accurate, bug 
free, etc.

> 3. I think my son uses Octave at UC Davis as a Matlab replacement with workable results. (He's subscribed to this list but I don't 
> know how much he reads it.)
>
> I'd love to know what is so compelling about Matlab that so many scientists won't switch over to (what to me is the dead-easy to 
> use) Python + tools.
>
> I don't actually know how I ended up here. I started off as a kernel hacker, writing drivers and other interrupt-savy code, 
> including assembly when needed. Now I'm like 14 levels of abstraction above that. Hell, I'm even writing web stuff now, which 
> (deep down) just doesn't seem nearly as hard core as writing device drivers. What a long strange trip it's been. If I'm still 
> programming at 65, I'll probably just actually be advising robots how to improve the UIs they write and how to write more 
> human-readable code. Er, robot-readable.

We were discussing today how to debug a gpio driven reset line problem for our camera on the new chipset with the new kernel version.

>
> - Joe 

sdw



More information about the FoRK mailing list