首页|Patent Issued for Machine learning training dataset optimization (USPTO 12046021 )
Patent Issued for Machine learning training dataset optimization (USPTO 12046021 )
扫码查看
点击上方二维码区域，可以放大扫码查看
原文链接
NETL
NSTL
Reporters obtained the following quote from the background information supplied by the inventors:“The invention relates to the field of machine learning.“To create accurate machine learning models, large quantities of properly-annota ted training data arerequired. For example, object recognition models require t raining sets of annotated images to enable themodel to learn robust classificat ion.“However, creating large sets of labeled images involves a costly and labor-inte nsive process, which istypically performed manually by annotation specialists. Moreover, the labeling resources and budgets arenot always allocated in an opti mal way, because it is not always known at the training stage which rawdata sam ples are potentially the most useful for training the model.
BusinessCyborgsDataloop Ltd.Emergi ng TechnologiesMachine Learning
NETL
NSTL
