Today we have a dedicated section for ethical issues. They may not directly related to data science, but I believe that we tech workers need to have a firm grasp of how our work is going to affect our people’s life, and make moral decisions. As Dwight D. Eisenhower put it, “A people that values its privileges above its principles soon loses both.” 😔
Competition on Adversarial Attacks & Defences
Registration for Competition on Adversarial Attacks & Defences（CAAD）is open now! https://t.co/cjVEXGQ6yI
— GeekPwn (@GeekPwn) May 31, 2018
Why Batch Norm Works
Timely paper from @ShibaniSan, Dimitris Tsipras, @andrew_ilyas , and @aleks_madry providing some new insights into why batch norm works. They perform a number of clever experiments to work it out, finding that internal covariate shift is a red herring! https://t.co/fJV4DjagW5 pic.twitter.com/G20yf9pMeJ
— Ari Morcos (@arimorcos) May 30, 2018
Ethics
My fellow engineers: grow a goddamn ethical backbone. "This is inevitable because someone else will do it" doesn't mean that *you* have to personally contribute to it, and if all of us decline, then it won't happen and isn't inevitable. https://t.co/TQWqBiltLa
— Liz Fong-Jones (@lizthegrey) May 30, 2018
“Algorithms may not care about equal employment opportunities, but our civil rights laws require it,” says lawyer in case filed after our reporting on employers who targeted Facebook job ads only to young workers. https://t.co/b6yEKNUdCK
— Julia Angwin (@JuliaAngwin) May 30, 2018
Here's an age discrimination lawsuit filed against IBM last week in Texas: https://t.co/LA9OHhPlQc
— Ariana Tobin (@Ariana_Tobin) May 31, 2018
Many many more examples of the company's marketing efforts toward Millennials, which was just a piece of our story (https://t.co/ACGiwPMjsa)
High Time to Regulate #FaceRecognition #AI @IntuitMachine https://t.co/q5dNDoh55Y pic.twitter.com/UZiLJeMDl8
— KDnuggets (@kdnuggets) May 31, 2018
These amazing students are working hard to make sure that AI is developed safely and responsibly https://t.co/R1edU9X5Da
— Ian Goodfellow (@goodfellow_ian) May 31, 2018
Data Visualization
ICYMI, 📊short-course & 📽s!
— Mara Averick (@dataandme) May 31, 2018
"Data Visualization Pitfalls to Avoid" 👩💻 @tamaramunzner https://t.co/r9v5FNqHMQ #dataviz #infovis #SoDS18 pic.twitter.com/8WpZgpGjXV
Population distribution of #Finland. #datavizhttps://t.co/IecAZe1pcs pic.twitter.com/VgeTWFBMT0
— Randy Olson (@randal_olson) May 31, 2018
I am making some edits to my slide on '3D Exploding Pie Charts are Bad' but WHAT THE HECK HAVE I CREATED?! What is this?! If there is a god, how could he allow such things to be created?
— Frank Elavsky ᴰᵃᵗᵃ ᵂᶦᶻᵃʳᵈ (@Frankly_Data) May 15, 2018
Worst #dataviz I've ever made. I hate this so much. pic.twitter.com/1xyyWNFmyc
Three ways of visualizing a graph on a map https://t.co/BRgzmzHu1E #rstats #DataScience
— R-bloggers (@Rbloggers) May 31, 2018
Notables
We've open sourced a @PyTorch implementation of our paper on disentangling discrete and continuous factors at https://t.co/gR3UzJtsJ2 pic.twitter.com/xWqtSDgApR
— Emilien Dupont (@emidup) May 31, 2018
Version 4.8 of the #HXL data library libhxl-python is now available on PyPi and GitHub, with many enhancements (especially for data validation).https://t.co/920rho52GHhttps://t.co/18UWXTEU4e#Python #humanitarian #data cc @humdata
— David Megginson (@david_megginson) May 31, 2018
Want to get notified when your fastai training is finished? Here's a little callback that does just that!https://t.co/sYqbBucpTo pic.twitter.com/uqgyWAeywd
— Jeremy Howard (@jeremyphoward) May 31, 2018
If your company trains machine learning models, check out https://t.co/UApm9VG0wM https://t.co/bF0zC0rYlh
— Peter Skomoroch (@peteskomoroch) May 31, 2018
Kumaraswamy distribution – a Beta-like distribution with a simple closed form of CDF
— ML Review (@ml_review) May 31, 2018
(The CDF for a beta distribution cannot be reduced to elementary functions unless its parameters are integers)https://t.co/7NaQScB7ev #MachineLearning pic.twitter.com/21okpupzgw
GANs
Improving CycleGAN for Image-to-Anime. Ongoing work from @_aixile. https://t.co/30sNtMSYGh
— hardmaru (@hardmaru) May 31, 2018
Happy to share our new work on MolGAN: An implicit generative model for small molecular graphs. GANs + RL for generating graphs with desired properties: https://t.co/rvsQ410HI9 - Some benefits over VAEs, but can still suffer from mode collapse. Joint work with @nicola_decao pic.twitter.com/bKcY7Tahdq
— Thomas Kipf (@thomaskipf) May 31, 2018
Causal Models
Turns out that with multiple causes, you don't really need strict ignorability or the "selection on observables" assumption! Break-through idea for #causalinference from @yixinwang_ and @blei_lab https://t.co/adQknuNV0c h/t @jakehofman pic.twitter.com/Vjyjn0etWf
— Amit Sharma (@amt_shrma) May 31, 2018
If you like this work, check out our causal models for genomics. This is where I first discovered this with @blei_lab, and also scaled it to huge GWAS data sets. https://t.co/CnKTFtxGBW https://t.co/oS39Dh5doy
— Dustin Tran (@dustinvtran) May 31, 2018
Algorithmic Trading in R
Algorithmic Trading: Using Quantopian’s Zipline Python Library In R And Backtest Optimizations https://t.co/Mfyz19cNmS #rstats #DataScience
— R-bloggers (@Rbloggers) May 31, 2018
Miscellaneous
Researchers created a series of images showing how different scenes might appear to various pets and pests. Human eyesight is roughly seven times sharper than a cat, 40-60x sharper than a rat or a goldfish, and 100x sharper than a fly or a mosquito. https://t.co/rc8DMQ3XAJ pic.twitter.com/m3GZM0Yxrq
— hardmaru (@hardmaru) May 31, 2018
Very interesting essay by Aaron Hertzmann about the role of technology in art and on the question of whether computers coukd be artists.
— Yann LeCun (@ylecun) May 31, 2018
Aaron argues that art is a social activity, and therefore machine can't really... https://t.co/c0N3Oq3NEi
The evolution of #datascience, #dataengineering, and #AI
— KDnuggets (@kdnuggets) May 31, 2018
The @OReillyMedia Data Show #Podcast: A special episode to mark the 100th episode @bigdata @pacoid https://t.co/QKvaL6CFX3 pic.twitter.com/By87CU0fZC
I used to complain about getting hit up by recruiters so often. A friend gently pointed out that this isn’t a good look. People in other fields (and underrepresented folks in ours) wish they had this “problem”. I was grateful for the nudge; perhaps you might be, too.
— jacobian (@jacobian) May 31, 2018
Put a 🔎 in your profile name if you're openly looking for a job so those of us that know of open roles can spot u! Also see @NewDataSciJobs! https://t.co/IVc1lavz1w
— Data Science Renee (@BecomingDataSci) June 13, 2017
Just re-formatted all the deep learning book contents I had previously posted (i.e., the appendix sections) in LaTeX. Was a bunch of work, but looks prettier and most importantly, it hopefully lowers the barrier for adding more & new content ;) https://t.co/v7oFwgi8ew
— Sebastian Raschka (@rasbt) June 1, 2018