首页|Patent Issued for Real-time predictions based on machine learning models (USPTO 12106199)
Patent Issued for Real-time predictions based on machine learning models (USPTO 12106199)
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
By a News Reporter-Staff News Editor at Robotics & Machine Learning Daily News Daily News – Salesforce Inc. (San Francisco, Califo rnia, United States) has been issued patent number 12106199, according to news r eporting originating out of Alexandria, Virginia, by NewsRx editors. The patent’s inventors are Bansal, Kaushal (Pleasanton, CA, US), Dasgupta, Amrit a (San Francisco, CA, US), Jagota, Arun Kumar (Sunnyvale, CA, US), Karanth, Rake sh Ganapathi (San Mateo, CA, US). This patent was filed on April 20, 2023 and was published online on October 1, 2 024. From the background information supplied by the inventors, news correspondents o btained the following quote: “Field of Art “This disclosure relates in general to machine learning based models, and in par ticular to performing real-time tasks using predictions using machine learning b ased models. “Description of the Related Art “Several online systems, for example, multi-tenant systems use machine learning based models for making predictions. These machine learning based models are inv oked by applications that may execute on client devices. Furthermore, for certai n applications, a multi-tenant system may generate scores using the machine lear ning based models on a periodic basis, for example, once every hour or once a da y. The multi-tenant system provides the results of execution of the machine lear ning based models to the client device. The multi-tenant system provides the gen erated scores to the users of the tenants for invoking via their applications. T his allows the use of powerful hardware of the multi-tenant system to execute th e machine learning based model while incurring low communication overhead while transmitting the generated results to the client devices. Such techniques are su ited for applications that do not require results of the machine learning models in real-time. However, such systems are inadequate if the results of execution of the machine learning based model are needed in real-time. For example, a clie nt device may not be able to generate accurate results immediately in response t o changes in the features used as input. The user is required to wait until the execution of the model is triggered on a periodic basis.
NETL
NSTL
