search

uchicago-computation-workshop / 2018_spring_conference

0 stars 1 forks source link

Reid McIlroy-Young #10

Open reidmcy opened 6 years ago

rickecon commented 6 years ago

Really strong project. Good presentation.

WZheng-94 commented 6 years ago

Interesting topic Can you name some social implication of this topic?

bensoltoff commented 6 years ago

Trying to visualize the model results. My understanding is that these deep neural nets are hard to visualize in an intuitive way. If you can figure out a good approach, that definitely sounds like a worthwhile package to develop for R.

jamesallenevans commented 6 years ago

Could be clearer about the precise question (as someone know your question) Nice trace of computational techniques used across sciences, social sciences and humanities Is "CS" what you want to capture...is it computational? Is it the journals that are classifying the work passively?

rodrigovaldes commented 6 years ago

Maybe you can relate your results to other data, like digital social networks, number of web pages, or something behind the general trends.

Btw, Idk if I will say that WoS is nice and clean.

reidmcy commented 6 years ago

@ZhengErWei I'm more interested in proving the model works and can be used by other people, but there is very little understanding of how scientists use computational tools. If we can better understanding of who is using the tools, we can improve the tools, and by getting better attribution encourage more people to create tools as right now there is not that many incentives to publish code.

reidmcy commented 6 years ago

@jamesallenevans I conflated the two in my presentation as that's a bit of a hard question to answer with the way the data are structured. The question is already a paragraph in my draft.

The journals are acting as passive classifiers, there is a non insignificant amount of messiness to the data.That the model works at all is actually very interesting. I'm working right now on getting a better system for generating training data, as luckily it only takes ~15 000 exposures to train the model.

reidmcy commented 6 years ago

@rodrigovaldes Yeah, it's better than most though.

I'm looking at using GitHub, but I won't have the data until the summer.

xiaow2 commented 6 years ago

Fancy fancy. I would like to know more about the working memory and long-term memory of neural nets. Thanks!