MA­CHINE LEARN­ING FOR EV­ERY­DAY DE­VEL­OP­ERS

Ama­zon SageMaker and AWS Deep­Lens.

HWM (Malaysia) - - SPECIAL -

From per­son­al­ized rec­om­men­da­tions to the cashier-free Ama­zon Go gro­cery store, there is a long her­itage of ma­chine learn­ing (ML) at Ama­zon. While there’s a lot of hype sur­round­ing ML and AI, it’s still very early days for most com­pa­nies, and skilled ML prac­ti­tion­ers are few and far be­tween. To re­move the heavy lifting from build­ing ML mod­els from scratch, AWS in­tro­duced Ama­zon SageMaker, which is de­signed for ev­ery­day de­vel­op­ers to eas­ily build, train, and de­ploy ML mod­els.

The fully man­aged ser­vice takes away the com­plex­ity of ML im­ple­men­ta­tion by pro­vid­ing 10 com­monly-used al­go­rithms, broad frame­work sup­port, as well as one-click train­ing at petabyte scale, and one-click tun­ing via hy­per-pa­ram­e­ter op­ti­miza­tion for high­est pos­si­ble ac­cu­racy. The re­sult­ing ML model can then be de­ployed on an auto-scal­ing clus­ter of EC2 in­stances across mul­ti­ple avail­abil­ity zones. In ad­di­tion, SageMaker also pro­vides oneclick in­fer­ence, na­tive A/B test­ing sup­port, and se­cure HTTPS end­point for high through­put and low la­tency pre­dic­tions.

To help de­vel­op­ers re­al­ize the po­ten­tial of ML in a more hands-on man­ner, AWS also launched Deep­Lens, which is the world’s first deep learn­ing-en­abled wire­less video cam­era. The cus­tom-de­signed hard­ware sports a HD video cam­era, and is ca­pa­ble of run­ning over 100 bil­lion deep learn­ing oper­a­tions per sec­ond. It also comes with sam­ple projects and pre­trained mod­els, and is fully pro­gram­mable with Lambda func­tions. When in­te­grated with SageMaker, de­vel­op­ers can run and de­ploy trained ML mod­els on the de­vice in real time.

Newspapers in English

Newspapers from Malaysia

© PressReader. All rights reserved.