Chen, Yunji, Tianshi Chen, Zhiwei Xu, Ninghui Sun, and Olivier Temam. "DianNao Family."Communications of the ACM 59.11 (2016): 105-12. Web.
Most computers nowadays have several cores to their processor so they can handle more jobs at once. For specialized tasks, a processor core can be substituted for a hardware accelerator, a piece of hardware that speeds up a very specific type of program. To the creators of the DianNao family of hardware accelerators, it seems natural that "As computer architectures evolve toward heterogeneous multi-cores composed of a mix of cores and hardware accelerators, designing hardware accelerators for ML techniques can simultaneously achieve high efficiency and broad application scope.” This new generation of hardware accelerators will be just as focused as previous hardware accelerators, but because they boost the speed of learning algorithms their applications are much broader. ML can be used for a lot of different things, and speeding it up naturally makes all of those things faster.
Most computers nowadays have several cores to their processor so they can handle more jobs at once. For specialized tasks, a processor core can be substituted for a hardware accelerator, a piece of hardware that speeds up a very specific type of program. To the creators of the DianNao family of hardware accelerators, it seems natural that "As computer architectures evolve toward heterogeneous multi-cores composed of a mix of cores and hardware accelerators, designing hardware accelerators for ML techniques can simultaneously achieve high efficiency and broad application scope.” This new generation of hardware accelerators will be just as focused as previous hardware accelerators, but because they boost the speed of learning algorithms their applications are much broader. ML can be used for a lot of different things, and speeding it up naturally makes all of those things faster.