公开(公告)号：US20230186096A1

公开(公告)日：2023-06-15

申请号：US18161479

申请日：2023-01-30

Applicant: Google LLC

Inventor： Mitchel Weintraub ,  Ananda Theertha Suresh ,  Ehsan Variani

IPC: G06N3/084 ,  G06N20/00 ,  G06F18/2431

CPC classification number: G06N3/084 ,  G06N20/00 ,  G06F18/2431

Abstract: Aspects of the present disclosure enable humanly-specified relationships to contribute to a mapping that enables compression of the output structure of a machine-learned model. An exponential model such as a maximum entropy model can leverage a machine-learned embedding and the mapping to produce a classification output. In such fashion, the feature discovery capabilities of machine-learned models (e.g., deep networks) can be synergistically combined with relationships developed based on human understanding of the structural nature of the problem to be solved, thereby enabling compression of model output structures without significant loss of accuracy. These compressed models provide improved applicability to “on device” or other resource-constrained scenarios.

公开(公告)号：US11568260B2

公开(公告)日：2023-01-31

申请号：US16654425

申请日：2019-10-16

Applicant: Google LLC

Inventor： Mitchel Weintraub ,  Ananda Theertha Suresh ,  Ehsan Variani

IPC: G06N20/00 ,  G06N99/00 ,  G06N3/08 ,  G06K9/62 ,  G06N20/10

Abstract: Aspects of the present disclosure enable humanly-specified relationships to contribute to a mapping that enables compression of the output structure of a machine-learned model. An exponential model such as a maximum entropy model can leverage a machine-learned embedding and the mapping to produce a classification output. In such fashion, the feature discovery capabilities of machine-learned models (e.g., deep networks) can be synergistically combined with relationships developed based on human understanding of the structural nature of the problem to be solved, thereby enabling compression of model output structures without significant loss of accuracy. These compressed models provide improved applicability to “on device” or other resource-constrained scenarios.

公开(公告)号：US11922322B2

公开(公告)日：2024-03-05

申请号：US18161479

申请日：2023-01-30

Applicant: Google LLC

Inventor： Mitchel Weintraub ,  Ananda Theertha Suresh ,  Ehsan Variani

IPC: G06N7/00 ,  G06F18/2431 ,  G06N3/084 ,  G06N20/00 ,  G06N20/10 ,  G06N5/00 ,  G06N10/00 ,  G06N10/80

CPC classification number: G06N3/084 ,  G06F18/2431 ,  G06N20/00

Abstract: Aspects of the present disclosure enable humanly-specified relationships to contribute to a mapping that enables compression of the output structure of a machine-learned model. An exponential model such as a maximum entropy model can leverage a machine-learned embedding and the mapping to produce a classification output. In such fashion, the feature discovery capabilities of machine-learned models (e.g., deep networks) can be synergistically combined with relationships developed based on human understanding of the structural nature of the problem to be solved, thereby enabling compression of model output structures without significant loss of accuracy. These compressed models provide improved applicability to “on device” or other resource-constrained scenarios.