datasci-v5   82

« earlier    

Where is the value in package peer review?
How is a package peer reviewer's time best spent?
When is the best time in a software package's life cycle to undertake peer review? A user-driven perspective.
R  datasci-v5 
11 weeks ago by mozzarella
Keep a Collection of Sparkly Data Resources • bowerbird
rOpenSci package that got me looking up Bowerbird architecture
R  statistics  datasci-v5 
11 weeks ago by mozzarella
Are random forests just a type of Monte Carlo?
brute force random variable generation over a fixed interval
vs.
bootstrap sampling with replacement
nice-thinking  datasci-v5 
november 2017 by mozzarella
Wealth inequality has been increasing for millennia - Daily chart
...because of agriculture? "The nomadic lifestyle is not conducive to wealth accumulation" I might've reworded the post to suggest 'the undervaluing of agriculture' being to blame for wealth inequality, if following that line of reasoning at all.
datasci-v5 
november 2017 by mozzarella
A Googler's Anti-Diversity Screed Reveals Tech's Rotten Core - The Atlantic
Products have been transformed into services given away “free” as an excuse to extract data from users. That data is woven into an invisible lattice of coercion and control—not to mention as a source of enormous profit when sold to advertisers or other interested parties. Apps and websites are designed for maximum compulsion, because more attention means more content, and more content means more data and thereby more value.
-- finding yet another description of the infor...
via:popular  nice-thinking  datasci-v5 
august 2017 by mozzarella
Apple removes apps used to bypass Chinese censors
this feels like a defeat
The Silicon Valley company has withdrawn “virtual private network” apps from the store, as it pulls all apps that do not comply with local law, even if the makers are based outside the country. 
datasci-v5 
july 2017 by mozzarella
Machine Learning Crash Course: Part 4 - The Bias-Variance Dilemma · ML@B
the overfit Fukushima earthquake model: a complex model was a tight fit to actual data, while the 'less accurate' linear regression line allowed more variance in data points. following the linear model's prediction, earthquakes of high magnitude would be expected to occur more frequently than in the overfit model - and thus more stringent safety measures likely would have followed.
statistics  machine-learning  datasci-V5  via:popular 
july 2017 by mozzarella
Research Blog: Using Deep Learning to Create Professional-Level Photographs
this is amazing, but 'professional photographers' aren't really the best arbiters of what a 'good' photograph is. Also, training on national parks binds the results to a naturally bland subject, no pun intended. While an amazing achievement, nothing shown here demonstrates ability beyond a photographer's assistant/digital tech adjusting settings to a client's tastes in Capture One Pro. Jon Rafman's 9 Eyes project comes to mind as something that produced interesting photographs, as does the idea to find a more rigorous panel of 'experts' (e.g. MoMA), or training the model on streets/different locations than national parks.
vis  datasci-v5  via:HN 
july 2017 by mozzarella

« earlier    

related tags

a16z  algorithmic-marketing  art-as-reference  blockchain  chess  computer-science  culture  currency  darknet-markets  ethereum  ethereum:dapps  fossil-fuel-firearms  great-websites  ikebana-ball  ikebana  investigations:ffl  investing  machine-learning  mozz-stats  mozz-v3  nice-code  nice-datasources  nice-thinking  pdpd.biz  politics  privacy  r  scientists  sixteen-bean-domain  sixteen-bean-soup  solidity  spatial  spatial:vis  standing-rock-research  statistics  statistics:bayesian  style  sublime-text  text-mining  trees  true-statisticians  trust-the-process  vis  wiki  wikisketch  wow  wyo-59 

Copy this bookmark:



description:


tags: