Title:
HEAD ARCHITECTURE FOR DEEP NEURAL NETWORK (DNN)
Document Type and Number:
WIPO Patent Application WO/2024/040601
Kind Code:
A1
Abstract:
A head of a DNN receives an OFM from a backbone network of the DNN. The head can partition the OFM into feature groups having same sizes. The head can further generate local tensors from the features group. To generate a local tensor from a feature group, the head may further partition the feature group into two subgroups, e.g., based on a splitting factor. The spatial sizes of the subgroups depend on the splitting factor. One subgroup can be converted into an attention tensor. The other subject can be converted into a value tensor, which may have the same size as the attention tensor. The attention tensor and value tensor are mixed to produce the local tensor. The local tensors of all the feature groups can be aggregated to form a global vector, which can be fed into a classifier to output one or more classification determined by the DNN.
Inventors:
YAO ANBANG (CN)
LI CHAO (CN)
CAI DONGQI (CN)
LIU XIAOLONG (CN)
SHAO WENJIAN (CN)
LI CHAO (CN)
CAI DONGQI (CN)
LIU XIAOLONG (CN)
SHAO WENJIAN (CN)
Application Number:
PCT/CN2022/115254
Publication Date:
February 29, 2024
Filing Date:
August 26, 2022
Export Citation:
Assignee:
INTEL CORP (US)
YAO ANBANG (CN)
LI CHAO (CN)
CAI DONGQI (CN)
LIU XIAOLONG (CN)
SHAO WENJIAN (CN)
YAO ANBANG (CN)
LI CHAO (CN)
CAI DONGQI (CN)
LIU XIAOLONG (CN)
SHAO WENJIAN (CN)
International Classes:
G06N3/08
Foreign References:
| CN112597941A | 2021-04-02 | |||
| CN113921041A | 2022-01-11 | |||
| CN114092773A | 2022-02-25 | |||
| CN112347908A | 2021-02-09 | |||
| CN113312989A | 2021-08-27 | |||
| US20200356842A1 | 2020-11-12 |
Attorney, Agent or Firm:
BEIJING EAST IP LTD. (CN)
Download PDF:
Previous Patent: SEMICONDUCTOR DEVICE AND METHOD FOR MANUFACTURING THE SAME
Next Patent: DISPLAY PANEL AND DISPLAY APPARATUS
Next Patent: DISPLAY PANEL AND DISPLAY APPARATUS