Using Less Data

In a previous post I indicated that operating on smaller datasets was one of the best ways to ensure productivity and rapid experimentation when doing data science. At ICML 2013, Nikos and I presented a general strategy for using less data that applies to a wide variety of problems. The idea was inspired by the class-imbalanced subsampling heuristic. This is a well-known trick amongst computationa

[kisa12012]

少年よ，データを捨てよ