Titlebook: Data Management in Machine Learning Systems; Matthias Boehm,Arun Kumar,Jun Yang Book 2019 Springer Nature Switzerland AG 2019

赞成你

refine

insurgent

Resource Heterogeneity and Elasticity,ve made it easier than ever to access computing resources. For example, a public cloud such as Amazon EC2 allows users to acquire a cluster on demand and pay only for its actual usage. There is a blossoming ecosystem of tools, libraries, and platforms for ML in the cloud and cluster computing settin

Torrid

Wallow

Conclusions,ed systems, high-performance computing, programming languages, and more. Motivated by (1) data-driven applications, (2) data-intensive workload characteristics, and (3) data systems support for ML, we took a data-centric view and reviewed approaches for integrating ML in data systems as well as data

insightful

AER

Execution Strategies, discuss interesting runtime techniques, including special runtime optimizers. Finally, in the context of deep learning, there is a trend toward exploiting accelerators such as GPUs, FPGAs, and custom ASICs for training and scoring, which is another specific hybrid execution strategy.

FLIT

雪白

defendant