wannesm

I agree this is not how you should evaluate. But it is not how things are done in (correctly executed) ML. What is described is a validation set, not a test set. The test set should not be involved in training or(hyper)parameter optimization at any time.

I can recommend http://jupyter-contrib-nbextensions.readthedocs.io/en/latest...

If you're interested in actual implementations, the website of the Probabilistic Programming Workshop offers an overview of actively developed languages:

http://probabilistic-programming.org/

The Springer-Papers and Elsevier-Mendeley acquisitions are indeed interesting because they've happened at about the same time. It seems that the big publishers want to get a better grip on paper search and/or paper reading habits.

At the University of Leuven (Belgium), from this year on, all first year engineering students learn how to program in Python. Not yet an official reference to it but professor Dutre who teaches the course mentions it already on twitter: https://twitter.com/philipdutre/status/279716437982064640 https://twitter.com/philipdutre/status/301760614416674816

Prince already supports the new page floats attributes. A copyright notice in the ACM style works for me with "float: bottom;". The w3 css3 draft [1] has some more examples, also about two column-wide figures.

[1] http://www.w3.org/TR/css3-gcpm/#page-floats

I love Latex but it (or more specifically TeX) is showing its age. It's perfect for writing a paper given a good template but writing and debugging these templates (.cls, .sty) is unnecessarily hard. For my job (in academia), I have to update such Latex templates on a monthly basis and always end up looking at the current improvements for Latex. This updating task involves not 'writing' in Latex but 'programming' in Latex. If you are used to modern programming languages then Tex is stubborn and hard, therefore, I always have the feeling there is improvement possible (I mean, Python, Javascript, C++, ... are all easier to debug). Although people are working hard on this and doing interesting work (i.e., LuaLatex, Latex3), I feel the underlying Tex language has had its time and a more drastic change might be necessary. Remember that Tex is designed for computers with an order of magnitude less resources.

An increasingly interesting Latex replacement is, maybe surprisingly, html in combination with css and javascript. With every update of our browser the inspection and debugging tools become more powerful and every time I can track layout and programming bugs a little bit faster. With the addition of more properties targeting paged media in css3 it now becomes possible to also create nicely looking pdfs starting from html. Prince (http://www.princexml.com/), for example, is ahead of the browsers for CSS paged media properties and outputs a pdf-file directly. But also typical features for which we praise Latex are becoming available:

- Mathematical formulae: http://www.mathjax.org/

- Bibliographies: http://citationstyles.org/

- Advanced hyphenation: http://code.google.com/p/hyphenator/

Most people only use the basic commands and don't care about the underlying engine. Therefore, Latex as a 'writing' language is not too attached to the Tex 'programming' language. Pandoc (http://johnmacfarlane.net/pandoc/) could be enough to translate such Latex code to another markup format and use another engine. (Academics tend to use advanced Latex macros only when in need of space ;-))

I have quite some math in my notes and like Gollum also for its integration with MathJax. I prefer the txt-based approach but a lot of math is more digestible when typeset properly so I switch between the txt and browser view. (This is of course applicable to any txt-to-html based approach)