Speeding up learning in machine learning by Rice University
Couple of research topics from Rice University computer scientists that can speedup the learning in machine learning alogirthms. One is on SLIDE or sub-linear deep learning engine which is claimed to be the first algorithm for training deep neural nets faster on CPUs than GPUs. The other one is MACH or Merged Average Classifiers via Hashing where training times are about 7-10 times faster, and memory footprints are 2-4 times smaller.