2018 DS/ML digest 25

2018 DS/ML digest 25

Posted by snakers41 on September 28, 2018

Buy me a coffeeBuy me a coffee

Become a PatronBecome a Patron


Datasets

NLP

  • Amazing triplet loss boilerplate;
  • Araneum Russicum PROCESSED corpus available via wget:
wget http://rusvectores.org/static/rus_araneum_maxicum.txt.gz

Market

  • Google becomes evil as it builds products for China;
  • And Amazon may also be considering a US national roll-out (‘3000 stores’!?) of its unmanned convenience store concept
  • Google’s dataset search. Ofc it does not work. Blog post - based on schema.org/Dataset markup - this is the reason it does not work;
  • Yet another lottery Google competition - I guess people will stack 1000+ models;
  • Rule based + RL bots for Starcraft 2 beat the game bots now;

Blog Posts

  • Fchollet post on code quality;
  • OpenCV 4 + Raspberry pi;
  • Stack Overflow is 10 years old;

Libraries / code / books

  • Yet another online book about ML;
  • How to visualize the decision trees;
  • Decision tree visualization library;