公开(公告)号：US20240160919A1

公开(公告)日：2024-05-16

申请号：US18380725

申请日：2023-10-17

Applicant: MEDIATEK INC.

Inventor： Po-Hsiang Yu ,  Hao Chen ,  Peng-Wen Chen ,  Cheng-Yu Yang

IPC: G06N3/08 ,  G06N3/0464

CPC classification number: G06N3/08 ,  G06N3/0464

Abstract: In aspects of the disclosure, a method, a system, and a computer-readable medium are provided. The method of building a kernel reparameterization for replacing a convolution-wise operation kernel in training of a neural network comprises selecting one or more blocks from tensor blocks and operations; and connecting the selected one or more blocks with the selected operations to build the kernel reparameterization. The kernel reparameterization has a dimension same as that of the convolution-wise operation kernel.

公开(公告)号：US20240160928A1

公开(公告)日：2024-05-16

申请号：US18506145

申请日：2023-11-10

Applicant: MEDIATEK INC.

Inventor： Po-Hsiang Yu ,  Hao Chen ,  Cheng-Yu Yang ,  Peng-Wen Chen

IPC: G06N3/08 ,  G06N3/0464

CPC classification number: G06N3/08 ,  G06N3/0464

Abstract: A method for enhancing kernel reparameterization of a non-linear machine learning model includes providing a predefined machine learning model, expanding a kernel of the predefined machine learning model with a non-linear network for convolution operation of the predefined machine learning model to generate the non-linear machine learning model, training the non-linear machine learning model, reparameterizing the non-linear network back to a kernel for convolution operation of the non-linear machine learning model to generate a reparameterized machine learning model, and deploying the reparameterized machine learning model to an edge device.

公开(公告)号：US20240161013A1

公开(公告)日：2024-05-16

申请号：US18501039

申请日：2023-11-03

Applicant: MEDIATEK INC.

Inventor： Cheng-Yu Yang ,  Hao Chen ,  Po-Hsiang Yu ,  Peng-Wen Chen

IPC: G06N20/00

CPC classification number: G06N20/00

Abstract: A reparameterization method for initializing a machine learning model includes initializing a prefix layer of a first low dimensional layer in the machine learning model and a postfix layer of the first low dimensional layer, inverting the prefix layer to generate an inverse prefix layer of the first low dimensional layer, inverting the postfix layer to generate an inverse postfix layer of the first low dimensional layer, combining the inverse prefix layer, the first low dimensional layer and the inverse postfix layer to form a high dimensional layer, generating parallel operation layers from the high dimensional layer, and assigning initial weights to the parallel operation layers.