|
DELETE CASCADE posted:what kind of hosed up PI doesn't let the main grad student on the project be first author? how else will he ever graduate 
|
#
?
Mar 27, 2020 19:04
|
|
|
|
| # ? May 24, 2024 23:22 |
|
|
DELETE CASCADE posted:what kind of hosed up PI doesn't let the main grad student on the project be first author? how else will he ever graduate you're.... you're being facetious right?
|
|
#
?
Mar 27, 2020 19:07
|
|
|
academics is like the mob, credit kicks up to the top and never [graduating | getting tenure | moving on from being a postdoc] shits downhill
|
|
#
?
Mar 27, 2020 19:09
|
|
|
it comes down to tenure policies and/or how machiavellian the professor in question is. some places are really lovely about granting tenure, and the only thing that matters is your authorial position on the work and the impact factor of the journal. first author on some Nature or Science papers? great, you're in. third author in some field-specific publications or, god forbid, conference presentations? hahaahaha no get the gently caress out. so in those places to save their own hides the professors will insist that they get first author credit on anything they can, which means pushing around the grad students. "you did the work in my lab so i'm automatically the first author" regardless of contribution. other places are much more realistic about how academic research works, and will weight a grad student's paper that you supervised the same as if you were the first author. this is great because you can let the student put their name first, which helps their career tremendously and is more reflective of reality anyway, knowing that your tenure committee won't just throw it away. it's much healthier and less stressful. a disadvantage of doing this is that, if you want to apply to a different university later and they subscribe to the lovely model, you might be a little screwed over unless you can convince the committee to look more deeply. so some professors still insist on being first author on everything anyway because they're relentless ladder-climbers. i was lucky enough to go to a grad school like the second example, so i had four first-authored papers when i graduated, and then got a TT position at a university that has a similar policy so i can pass it on. e: once you're a tenured professor at a cushy institution like stanford, though, there's no excuse for continuing to assign yourself first author other than ego. Sagebrush fucked around with this message at 19:45 on Mar 27, 2020 |
|
#
?
Mar 27, 2020 19:41
|
|
|
are you suggesting that the tenured faculty at "elite" institutions are anything but the epitome of humility?
|
|
#
?
Mar 27, 2020 21:40
|
|
|
some very exciting research: https://distill.pub/2020/circuits/zoom-in/ turns out it's not too hard to figure out how artificial neural networks work if you're willing to put the time in
|
|
#
?
Apr 29, 2020 00:36
|
|
|
that’s certainly a good formalization of things
|
|
#
?
Apr 29, 2020 01:12
|
|
|
animist posted:if you're willing to put the time in found the problem. if neural nets don't give me results easily and with no thought on my part, what's the point??
|
|
#
?
Apr 29, 2020 01:46
|
|
|
animist posted:some very exciting research: https://distill.pub/2020/circuits/zoom-in/ are they comparing this to development of the microscope and the study of the atom sincerely?
|
|
#
?
Apr 29, 2020 06:39
|
|
|
quote:Focusing on the study of circuits, is universality really necessary? Unlike the first two claims, it wouldn’t be completely fatal to circuits research if this claim turned out to be false. But it does greatly inform what kind of research makes sense. We introduced circuits as a kind of “cellular biology of deep learning.” But imagine a world where every species had cells with a completely different set of organelles and proteins. Would it still make sense to study cells in general, or would we limit ourselves to the narrow study of a few kinds of particularly important species of cells? yeaaah, this sounds way too optimistic medical sciences can’t even vaccinate our pets, let alone all the nice animals in the food markets of Wuhan that give us sars outbreaks every couple of years... even organ donation between two people is often problematic and totally impossible, for example, between a whale and your mother, despite the superficial physical similarity Max Facetime fucked around with this message at 16:48 on May 2, 2020 |
|
#
?
May 2, 2020 10:55
|
|
|
was not expecting that post to end in a dunk
|
|
#
?
May 2, 2020 17:51
|
|
|
https://twitter.com/Strife212/status/1255789106522656773 https://twitter.com/Indoorsness/status/1256177588705169408
|
|
#
?
May 3, 2020 15:57
|
|
|
Vomik posted:are they comparing this to development of the microscope and the study of the atom sincerely? theyre overenthusiastic "disruption" dweebs but the work is still interesting imo Max Facetime posted:yeaaah, this sounds way too optimistic yeah i doubt this is going to have practical implications for a while. it helps satisfy my burning urge to understand what the gently caress neural nets are doing though then again, maybe it's reading tea leaves and will turn out to be a dead end. who knows??
|
|
#
?
May 3, 2020 16:19
|
|
|
gradient descent means literally into hell.
|
|
#
?
May 3, 2020 21:16
|
|
|
and stochastic gradient descent sometimes takes you to a place thats even worse
|
|
#
?
May 4, 2020 03:51
|
|
|
realized today that in the future there will be two kinds of talking machines - the ones that talk like wikipedia and the ones that talk like reddit
|
|
#
?
May 7, 2020 21:36
|
|
|
suffix posted:realized today that in the future there will be two kinds of talking machines - the ones that talk like wikipedia and the ones that talk like reddit the latter will be used for customer service, so that any woman who calls the cable company gets told she's ugly and gonna get raped machine learning solutions bringing reddit-grade discourse to the world of commerce
|
|
#
?
May 7, 2020 21:50
|
|
|
https://twitter.com/ryxcommar/status/1259289338854154240
|
|
#
?
May 10, 2020 02:17
|
|
|
prefabulate the amulite to reduce the chance of sinusoidal depleneration.
|
|
#
?
May 10, 2020 02:40
|
|
|
I remember when I did that last one by accident in college at my student job and got in trouble because we had to claw back and reprint a bunch of reports
|
|
#
?
May 10, 2020 04:09
|
|
|
If you rly think about it, life is just one big regression line.
|
|
#
?
May 10, 2020 06:25
|
|
|
most religion is just an argument about the direction the line is going
|
|
#
?
May 10, 2020 09:14
|
|
|
what could you do with a lot of real estate data
|
|
#
?
Jun 5, 2020 12:30
|
|
|
make a very solid argument for Maoism probably e: real talk depends what you're looking for. throwing some scikit-learn clustering/regression or xgboost at it prolly wouldn't hurt. T-SNE can be good for visualizing how your data is clustered too, interactive visualizations will give you the best sense of how it works animist fucked around with this message at 13:19 on Jun 5, 2020 |
|
#
?
Jun 5, 2020 13:13
|
|
|
maybe I should learn about what those things are gf works for news website. shes being given loads of real estate data, but what she needs to do is direct what to do with it. the crux is, what can be gained from analysing it? is this a problem of imagination, ie, we just need to think more about the potentials or a problem where the solution is throw the kinds of thing you're talking about at it, and make the data tell us what's relevant within it
|
|
#
?
Jun 5, 2020 13:32
|
|
|
when you're given a bunch of data, you wanna ask yourself "what's interesting about this setting" do the data span an interesting time period? are there policy variations over time or between jurisdictions or observations, what do you know about the variation, what kind of data do you have, and so on you can't just throw a dataset into an algorithm and have it come back with "hey actually here are the interesting facts and patterns", knowing what to look for is still a human skill
|
|
#
?
Jun 5, 2020 15:57
|
|
|
Exploratory data analysis is a thing, but it has to be guided by a question or at least a good sense of what value can come out of looking at the data.
|
|
#
?
Jun 5, 2020 15:59
|
|
|
Pinterest Mom posted:when you're given a bunch of data, you wanna ask yourself "what's interesting about this setting" also tbh if you're trying to understand data I'd reach for visualizations wayyy before ML because ML will just spit black boxes back at you. that's because I like to understand things tho
|
|
#
?
Jun 5, 2020 18:59
|
|
|
that t-sne looks interesting but I fell asleep looking at a YouTube of what it is Pinterest Mom posted:when you're given a bunch of data, you wanna ask yourself "what's interesting about this setting" hmm well this seems very reasonable. thanks
|
|
#
?
Jun 5, 2020 21:15
|
|
|
animist posted:also tbh if you're trying to understand data I'd reach for visualizations wayyy before ML because ML will just spit black boxes back at you. that's because I like to understand things tho actually I think you’ll find ml has a lot of trouble with black boxes
|
|
#
?
Jun 5, 2020 21:17
|
|
|
animist posted:also tbh if you're trying to understand data I'd reach for visualizations wayyy before ML because ML will just spit black boxes back at you. that's because I like to understand things tho sounds like you don't have the right mindset for doing ml
|
|
#
?
Jun 6, 2020 00:00
|
|
|
purpose of machine learning is to avoid understanding
|
|
#
?
Jun 6, 2020 01:38
|
|
|
redleader posted:sounds like you don't have the right mindset for doing ml few more months for my degree and then im never touching a fuckign gradient again echinopsis posted:that t-sne looks interesting but I fell asleep looking at a YouTube of what it is it tries to reduce the number of dimensions in the data while keeping the distances between all pairs of points the same. this is impossible but you can do a decent approximation with a bit of calculus and elbow grease. (fun fact, pretty much all of ml boils down to "dimensionality reduction with a bit of calculus and elbow grease")
|
|
#
?
Jun 6, 2020 07:29
|
|
|
come to think of it tho tsne needs data encoded as a pairwise distances so a lotta your raw real estate data (names, other text) prolly won't fit into it without some extra wrangling. you'd need to define some sorta distance metric... idk I don't work much with textual data sorry
|
|
#
?
Jun 6, 2020 07:34
|
|
|
neither lol
|
|
#
?
Jun 6, 2020 08:10
|
|
|
obviously just shove it into tf-idf vectors 
|
|
#
?
Jun 6, 2020 13:50
|
|
|
Hamming distance.
|
|
#
?
Jun 6, 2020 16:21
|
|
|
what does the number of processing steps required to turn pigs into food have to do with this
|
|
#
?
Jun 7, 2020 16:03
|
|
|
geeze, i know cops are bad but i thought we were doing the rich first
|
|
#
?
Jun 7, 2020 19:33
|
|
|
|
| # ? May 24, 2024 23:22 |
|
|
code:
|
|
#
?
Jun 8, 2020 15:12
|
|
