18380581. DATA PROCESSING METHOD AND RELATED DEVICE simplified abstract (Huawei Technologies Co., Ltd.)
Contents
DATA PROCESSING METHOD AND RELATED DEVICE
Organization Name
Inventor(s)
DATA PROCESSING METHOD AND RELATED DEVICE - A simplified explanation of the abstract
This abstract first appeared for US patent application 18380581 titled 'DATA PROCESSING METHOD AND RELATED DEVICE
Simplified Explanation
The data processing method described in the patent application involves obtaining two embedding vectors: one for a known data unit and its position, and another for the position of a to-be-predicted data unit. These embedding vectors are then processed using a target encoder to generate an output vector. The output vector, along with the second embedding vector, is further processed using a target prediction network to obtain the predicted data unit.
Key points of the patent/application:
- The method eliminates the need for setting additional position information as input for the target encoder.
- The quantity of latent variables in the intermediate output of the target encoder matches the quantity of input embedding vectors, reducing computation and memory consumption.
- The target encoder and target prediction network work together to accurately predict the to-be-predicted data unit.
Potential applications of this technology:
- Natural language processing: Predicting the next word or phrase in a sentence.
- Image recognition: Predicting the next pixel or feature in an image.
- Time series analysis: Predicting the next data point in a sequence.
Problems solved by this technology:
- Reduces the computational burden and memory consumption of the target encoder.
- Simplifies the input requirements by eliminating the need for additional position information.
- Improves the accuracy of predictions by incorporating both known data unit and position information.
Benefits of this technology:
- Increased efficiency in data processing.
- Improved accuracy in predicting future data units.
- Simplified input requirements for the target encoder.
Original Abstract Submitted
A data processing method includes: obtaining a first embedding vector for indicating a known data unit and a position of the known data unit and a second embedding vector for indicating a position of a to-be-predicted data unit; processing the first embedding vector by using a target encoder, to obtain an output vector; and processing the output vector and the second embedding vector by using a target prediction network, to obtain a to-be-predicted data unit. According to the method, M pieces of additional position information do not need to be separately set as input of the target encoder, and a quantity of latent variables of intermediate output of the target encoder is also consistent with a quantity of input embedding vectors, thereby reducing a computation amount and memory consumption of the target encoder.
