NIPS 2017 Notes [pdf]

https://cs.brown.edu/~dabel/blog/posts/misc/nips_2017.pdf

by stablemap • 9 years ago

240 24 9 years ago

cs.brown.edu

fredmonroe• 9 years ago

thanks for doing these, really appreciated

formalsystem• 9 years ago

Wow can't appreciate this enough. You put this together really quickly!

hackandtrip• 9 years ago

Wonderful content, appreciate those notes over usual videos that come out, really more usable. Is there a place where wonderful people usually share notes on conference like those?

Dangeranger3• 9 years ago

It makes me laugh thinking that these notes were probably written in LaTex and then exported.

The formatting for this overview is pretty great though. PDF with a TOC makes a very nice means of navigation.

castle-bravo• 9 years ago

It may have been done in org-mode and then exported to LaTeX. That's how I'd do it, anyway.

pdxdabel3• 9 years ago

(David/author here): I haven't actually seen org-mode before. I'll take a look!

I wrote these in sharelatex using a template and commands file I put together that help write notes quickly in latex (they're available here if anyone is interested: https://github.com/david-abel/latex_docs).

Cyph0n• 9 years ago

Is "exported" the correct term to use here? I would say the notes were written in LaTeX, and then compiled (in the software compiler sense) to a PDF.

Dangeranger3• 9 years ago

You are likely correct, exported is the wrong word, and my mind was within my personal org-mode workflow.

AlexCoventry• 9 years ago

I was guessing org-mode -> LaTeX -> PDF. org-mode is much better for taking notes.

emmelaich• 9 years ago

NIPS stands for "Neural Information Processing Systems"

(took me a while but I found it!)

daferna• 9 years ago

Somebody get this guy a PhD!

tuxguy• 9 years ago

these are awesome !!! thanks a bunch David, @stablemap !

pdxdabel3• 9 years ago

Hi all! David (the author) here -- first, thanks for the kind words and thanks to @stablemap for sharing! I'm glad folks have found them useful.

I'm planning on putting these together to future conferences I attend, too (I'm currently a 3rd year Ph.D student so I should be attending a few in the coming years).

mi_lk• 9 years ago

Hi David, very impressive of your notes, especially in this knowledge-intense machine learning conference, the way you organize diverse topics and have short summaries about them seems like a very useful skill to have. Do you have any methods/tips that you can share about your note-taking skills?

pdxdabel3• 9 years ago

Investing in a particular set of tools has been really critical for me -- once I found that I liked the flexibility and power of latex, I put effort into simplifying the process of using it.

So, for instance, I made some macros for commands that I use frequently, and a put together a latex template that includes the packages and basic structure of documents I write. Otherwise, just practice! I forced myself to write notes during class in latex when I took information theory earlier in grad school and that helped a lot.

Hope this helps. If I think of anything else I'll be sure to come back and post!

lustig• 9 years ago

Hi David! Thanks for a great summary! I will keep an eye on your blog for notes from upcoming conferences.

DyslexicAtheist2• 9 years ago

The event has also been compared to a dumpster fire[0] due to "sexualized events & speeches", is this guy exaggerating or is it safe to go there for women? Pretty off-putting tbh

[0] https://twitter.com/haldaume3/status/939910697911066624

yamaneko• 9 years ago

Smitha Milli gave a bit more of context about what happened. Link to her thread:

- https://twitter.com/SmithaMilli/status/940012716797739008

DyslexicAtheist2• 9 years ago

thanks, rather unspectacular then.

briga• 9 years ago

This seems like a pretty harmless joke to me. I have difficulty imagining why anyone would be offended by this, unless there's more to the story

lemondrops• 9 years ago

Also, this? https://twitter.com/SamiraEKahou/status/940085885101412352

I don't get why anyone but the wife herself should be offended by this.

eli_gottlieb• 9 years ago

It's interesting that when asked about why we use hierarchical methods in RL, ML, and cogsci, people haven't talked about the circuit complexity and information-theoretic reasons for using hierarchies. IIRC, a "deeper" circuit can represent a given function with exponentially fewer units/gates than a "shallower" circuit, and hierarchical methods also narrow down only the predictive information in the (supervised) dataset, according to information-bottleneck methods.

NIPS 2017 Notes [pdf]

Related Stories