Hacker News new | past | comments | ask | show | jobs | submit login
NIPS 2017 Notes [pdf] (brown.edu)
240 points by stablemap on Dec 12, 2017 | hide | past | favorite | 24 comments



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).


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?


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!


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


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.


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?


NIPS stands for "Neural Information Processing Systems"

(took me a while but I found it!)


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.


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.


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


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


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


(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).


Gary Schwartz[0] and Howard Abrams[1][2] both have nice explainers about org-mode which explain the advantages and how to get started.

There's also a nice discussion on the Emacs Reddit page[3] about "Why Org-Mode?"

[0] https://harryrschwartz.com/2016/04/29/getting-started-with-o...

[1] http://www.howardism.org/Technical/Emacs/learning-org-mode.h...

[2] http://www.howardism.org/Technical/Emacs/journaling-org.html

[3] https://www.reddit.com/r/emacs/comments/3nvmvr/why_orgmode/


Somebody get this guy a PhD!


thanks for doing these, really appreciated


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


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


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


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

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


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


Well, according to the notes there was an entire talk about bias where the example of a Turkish sentence that didn't contain gendered pronouns "X is a nurse, X is a doctor" gets translated to "She is a nurse, he is a doctor" by Google Translate. And apparently that's a problem that needs fixing rather than a reasonable translation, although we all know that Google Translate is trained on the work of human translators, where it presumably learned this (demographically correct) bias from. So I guess they need fixing too.

Seems like NIPS might have a systematic issue with hyper-sensitivity to perceived slights. But it's the intersection of academia and SV companies so is it really a surprise?


thanks, rather unspectacular then.


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

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




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: