Unfortunately, many machine learning algorithms aren't easily parallelizable. Consider any boosting ensemble learner, such as the popular random forest, which optimizes the result from the previous iteration. (I run the machine learning startup diffbot.com and we could really use faster cycles.)
Comments
Unfortunately, many machine learning algorithms aren't easily parallelizable. Consider any boosting ensemble learner, such as the popular random forest, which optimizes the result from the previous iteration. (I run the machine learning startup diffbot.com and we could really use faster cycles.)
what sort of algorithms would you like to have faster variants/implementations of?