Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority claimed from PCT/CN2020/108381external-prioritypatent/WO2022032471A1/en
Application filed by Univ Chinese Hong Kong ShenzhenfiledCriticalUniv Chinese Hong Kong Shenzhen
Publication of ZA202303562BpublicationCriticalpatent/ZA202303562B/en
The present invention is applicable to the technical field of model training. Provided are a neural network model training method, device and system. The method includes: acquiring an original data set and training an original neural network model according to the original data set; identifying a noise label from the original neural network model; and modifying the noise label and training a new neural network model according to a modified data set. The present invention is performed by training an original neural network model from an original data set first, and identifying a noise label in the original neural network model, thereby determining an erroneous label in the original data set, after modifying the erroneous label, finally training a new neural network model according to a modified data set, thus has high accuracy and good interpretability as the erroneous label is directly determined from the network model and modified, so that a new neural network model obtained through final training has a better anti-interference effect.
ZA2023/03562A2018-10-022023-03-13Method, device, storage medium and apparatus of training a neural network model
ZA202303562B
(en)
DEVICE AND METHOD FOR COMPUTING A FINGERPRINT OF AN AUDIO SIGNAL, DEVICE AND METHOD FOR SYNCHRONIZING AND DEVICE AND METHOD FOR CHARACTERIZING A TEST AUDIO SIGNAL