公开(公告)号：US11757703B1

公开(公告)日：2023-09-12

申请号：US17937427

申请日：2022-09-30

Applicant: Amazon Technologies, Inc.

Inventor： Eric Ray Hotinger ,  Kathiravan Kalimuthu ,  Arvind Jayasundar ,  Chao Duan ,  Ippokratis Pandis ,  Hitenkumar Sonani ,  Davide Pagano ,  Yousuf Hussain Syed Mohammad ,  Bruce William McGaughy ,  Bin Zhang

IPC: H04L41/0663 ,  H04L9/08 ,  H04L41/0604

CPC classification number: H04L41/0663 ,  H04L9/0825 ,  H04L41/0627

Abstract: A database service may distribute resources across different geographic locations or other infrastructures to increase availability of the resources and may provide multiple locations to access resources and isolate failure of resources to a respective location or infrastructure. The processing resources in differing fault tolerance zones may be able to continue operating in the event of an outage impacting an entire fault tolerance zone. The database service may generate a supporting processing cluster in the differing fault tolerance zone that handles at least a portion of the access requests of an initial processing cluster. The database service may provision the supporting processing cluster in a separate fault tolerance zone that has a similar capacity and may provision and maintain the cluster in order to preclude the potential of not having sufficient capacity to recover upon failure of a single fault tolerance zone.

公开(公告)号：US11481906B1

公开(公告)日：2022-10-25

申请号：US16370723

申请日：2019-03-29

Applicant: Amazon Technologies, Inc.

Inventor： Hareesh Lakshmi Narayanan ,  Rahul Sharma ,  Arvind Jayasundar ,  Vikram Madan

IPC: G06T7/187 ,  G06K9/62 ,  G06N20/00 ,  G06V30/414

Abstract: Techniques for active learning-based data labeling are described. An active learning-based data labeling service enables a user to build and manage large, high accuracy datasets for use in various machine learning systems. Machine learning may be used to automate annotation and management of the datasets, increasing efficiency of labeling tasks and reducing the time required to perform labeling. Embodiments utilize active learning techniques to reduce the amount of a dataset that requires manual labeling. As subsets of the dataset are labeled, this label data is used to train a model which can then identify additional objects in the dataset without manual intervention. The process may continue iteratively until the model converges. This enables a dataset to be labeled without requiring each item in the data set to be individually and manually labeled by human labelers.