In software, naming matters, because names reflect how you think about a problem. Code is also communication, and naming is a big part of making it work.
— François Chollet (@fchollet) September 7, 2018
In software, naming matters, because names reflect how you think about a problem. Code is also communication, and naming is a big part of making it work.
— François Chollet (@fchollet) September 7, 2018
The most under-appreciated programming language of all time is probably SQL. SQL is (generally) well-designed and omnipresent, but it's almost never brought up as an example of how to design a good language.
— John Myles White (@johnmyleswhite) September 5, 2018
This has been our experience running Kaggle as a remote-first company. When you orient the whole team around being remote, it works phenomenally well https://t.co/NiPF10zCpS
— Ben Hamner (@benhamner) September 4, 2018
The reality was a lot different of course. Human passions and jealousies drive much more of science than is commonly understood, and "scientific truths" are far fuzzier than you might think.
— Bharath Ramsundar (@rbhar90) September 2, 2018
"Narrowness of experience leads to narrowness of imagination."
— michael_nielsen (@michael_nielsen) September 2, 2018
Reminds me of Rob Pike's wonderful talk "Systems Software Research is Irrelevant", especially of this bit: https://t.co/SqEAhnhLUw pic.twitter.com/ED1ygY2Imf
— michael_nielsen (@michael_nielsen) September 2, 2018
One of the best pieces of advice I've received: Don't do a PhD that isn't fully funded. https://t.co/BWM7D2LiAg
— Chris Albon (@chrisalbon) September 1, 2018
Good thread. As a first-gen grad student, I thought you had to take major loans for a CS PhD. I'm lucky to have had great friends (whose parents had advanced degrees) to (i) encourage me to go to grad school and (ii) tell me I'd get paid to do so! https://t.co/hJ9xhyPcet
— Eric Rozner (@erozner) September 1, 2018
It took me a while to realize that data analyses are not naturally occurring phenomena. They must be constructed by people, for people, and hence, designed appropriately. https://t.co/TQ8ViD7NaR
— Roger D. Peng (@rdpeng) August 31, 2018
I find character-level language modeling much more simpler and elegant compared to using traditional word token pipelines. I think it is also a more interesting research problem to study languages at the character level compared to word level.
— hardmaru (@hardmaru) August 30, 2018
Double down on automation
— boB Rudis (@hrbrmstr) August 29, 2018
BUT
Don't build purely automated tools
Use Human-Computer symbiosis as an architecture to make defenders "bionic"
(I usually say 'smarter-better-faster' since kids tdy have never seen The Bionic Man :-)#certds18
I think notebooks encourage putting everything into a single document, but that goes against clear communication and developing a coherent narrative of an analysis. https://t.co/bYkpED02Gk
— Roger D. Peng (@rdpeng) August 29, 2018