filmov
tv
beebucket mask detector
Показать описание
What are reasons, which often prevent usage of Machine Learning and Artificial Intelligence at the networks edge, close to the data originators?
Architectural Cohesion: There are many existing offerings based on centralized ML/AI services. Their concept is based on data centralization. Such solutions are dependent on connectivity to the services.
Connectivity, Capacity and Costs: Many use-cases are dealing with highly dense data, such as video, or audio. Centralizing such data from many data originators is often not a viable option at larger scale, simply because of implied cost vs. anticipated business value.
Privacy: Not every use case permits data exposure outside the data originators premise.
Resources: Distributed scalable virtual resources are not a given, especially at the networks edge, or they may be considered too costly to withstand a value assessment.
How about minimizing the required resource allocation of ML/AI at the networks edge so drastically, that it can run on resource restricted environments?
Architectural Cohesion: There are many existing offerings based on centralized ML/AI services. Their concept is based on data centralization. Such solutions are dependent on connectivity to the services.
Connectivity, Capacity and Costs: Many use-cases are dealing with highly dense data, such as video, or audio. Centralizing such data from many data originators is often not a viable option at larger scale, simply because of implied cost vs. anticipated business value.
Privacy: Not every use case permits data exposure outside the data originators premise.
Resources: Distributed scalable virtual resources are not a given, especially at the networks edge, or they may be considered too costly to withstand a value assessment.
How about minimizing the required resource allocation of ML/AI at the networks edge so drastically, that it can run on resource restricted environments?