Title:
INFORMATION PROCESSING APPARATUS, INFORMATION PROCESSING METHOD, NON-TRANSITORY COMPUTER READABLE MEDIUM
Document Type and Number:
WIPO Patent Application WO/2021/100121
Kind Code:
A1
Abstract:
A Model training system (100) includes An ANN model trainer means (101) for training an ANN model using training data, an Information matrix computation means (102) for computing information matrix, which implies the importance of ANN parameters, from training information, and a Policy model trainer means (103) for training traditional light-weight machine learning (non-DL) policy model using the training data and the information from the information matrix. Accordingly, the policy model can generate policy that indicates the important ANN parameters for omitting some inference computation of the ANN model.
Inventors:
SOMBATSIRI SALITA (JP)
Application Number:
PCT/JP2019/045281
Publication Date:
May 27, 2021
Filing Date:
November 19, 2019
Export Citation:
Assignee:
NEC CORP (JP)
International Classes:
G06N3/08; G06N3/04
Foreign References:
| US20180046898A1 | 2018-02-15 |
Other References:
UEYOSHI, KODAI ET AL.: "IEICE Technical Report", A METHOD TO CALCULATE DNN EFFICIENTLY BY PREDICTING INVALID NEURONS, vol. 119, no. 18, 2 May 2019 (2019-05-02), pages 97 - 102
Attorney, Agent or Firm:
IEIRI Takeshi (JP)
Download PDF:
Previous Patent: HAIR MEASURING METHOD AND HAIR MEASURING SYSTEM
Next Patent: DESIGN ASSISTANCE SYSTEM AND DESIGN ASSISTANCE PROGRAM
Next Patent: DESIGN ASSISTANCE SYSTEM AND DESIGN ASSISTANCE PROGRAM