Python and Astronomy / PyCon 2017 Keynote

[freiform]

Hi,

a friend forwared me this video of Jake Vanderplas' keynote on this years PyCon. It may not hold new information about astronomy itself, bit I found it very interesting nontheless.

 

Sven

 

[Martin Meredith]

Nice video. I relatively recently (in the scheme of things) converted from Matlab to Python and never looked back. I also use his book extensively for teaching Python Data Science (also freely available as a bunch of Jupyter Notebooks on the web). Definitely the way to go for developing astronomy apps.

Martin

[freiform]

I cannot quite make the (complete) switch from Matlab to Python. I've been using Matlab daily for nearly 11 years now, and it is kind of my goto-response when I want to do something very quickly. I try to assign certain projects and tasks as Python-only. Some are by definition, because sometimes Matlab is just, well, Matlab. Maybe I'll get there someday.

Sven

[Paul M]

It was an interesting video but I  know very little about Python. Actually my ignorance is such that I actually thought that Python was developed for the Raspberry Pi. Yes, I know, I know 

Py=Pi. No? Easy mistake!

[Helen]

Thanks for this - bookmarked to watch!  I'm just starting a PhD analysing Comet 67P data and am working my way up the Python learning curve.  I'm trying to be disciplined and learn it properly rather than just botch something together for what I need to do at any one time!  If anyone has particular recommendations for learning resources for data reduction etc I'd be very interested!!

Thanks

Helen

[Martin Meredith]

12 hours ago, Helen said:

Thanks for this - bookmarked to watch!  I'm just starting a PhD analysing Comet 67P data and am working my way up the Python learning curve.  I'm trying to be disciplined and learn it properly rather than just botch something together for what I need to do at any one time!  If anyone has particular recommendations for learning resources for data reduction etc I'd be very interested!!

Thanks

 

Helen

Assuming you haven't already done so, the best starting point is to download your python + data science packages from  https://www.anaconda.com. Using Jupyter notebooks is highly recommended. Then there is Jake's handbook online here

https://jakevdp.github.io/PythonDataScienceHandbook/

For image processing, check out Pillow

https://pillow.readthedocs.io/en/4.3.x/

There's much more out there but these might be useful (not sure how far into Python you are yet!).

Martin

 

[Martin Meredith]

12 hours ago, freiform said:

I cannot quite make the (complete) switch from Matlab to Python. I've been using Matlab daily for nearly 11 years now, and it is kind of my goto-response when I want to do something very quickly. I try to assign certain projects and tasks as Python-only. Some are by definition, because sometimes Matlab is just, well, Matlab. Maybe I'll get there someday.

Sven

I thought the same (and speaking as a daily Matlab user for the last 20 years) but I really regret not switching completely to Python 5 years ago. I work in signal processing, and to my surprise I found Python to be a little faster than Matlab too even for areas where you'd expect Matlab to have the edge (e.g. filtering). I find my Python code to be about a third as long as the equivalent Matlab code.

Matlab will linger on for a while longer but I really doubt there will be much serious new development in that language. The area where Python appears at first glance deficient wrt Matlab is the ease of building interfaces, but I've recently started to use Kivy in anger, and the GUIs are far superior to what it possible in Matlab (+platform-independent).

Martin

[Macavity]

Interesting stuff. I suppose I am lucky in "knowing" that:
"Python is a High Level Computer Language" (roughly) as
a start...  Good Talk on it's various virtues, though? 

[rotatux]

12 minutes ago, Martin Meredith said:

For image processing, check out Pillow

I didn't check v4 but at least up to v3 it only handles 8-bit images, so that may be a severe limitation for you.

OpenCV should be more flexible, it's more oriented to computer vision as name says, but has strong general image processing features, and (the killer feature) manipulates images as NumPy arrays.

[Martin Meredith]

This is what Pillow v4.3 has

Agree, OpenCV worth a look too. (BTW it is pretty straightforward to treat Pillow images as NumPy too with a relatively minor overhead https://stackoverflow.com/questions/384759/pil-and-numpy)

Martin

[Paul M]

9 hours ago, Macavity said:

"Python is a High Level Computer Language"... 

Whoa there! Back up, I'm lost already!

[Macavity]

10 hours ago, Paul M said:

Whoa there! Back up, I'm lost already!

Heheh. Long time since I did any "professional" programming! I get the impression
Python might be able to talk to an Arduino via simple dialogue boxes from Windows?
I never *really* solved synchronisation of such stuff from Windows Visual studio.

Thanks for this anyway! P.S. Pycon, Eh? I'd rather watch from the safety of SGL! 
http://www.motherjones.com/politics/2013/03/pycon-2013-sexism-dongle-richards/