Neural network training method and apparatus using trend
AssigneeSamsung Electronics Co., Ltd.
InventorsJUN HAENG LEE, SEONGJIN PARK
Life of this patent
Estimating the term…
Abstract
A neural network training method and apparatus are disclosed, where the neural network training method using trend includes receiving original training data and trend information, obtaining a first prediction value and a first loss from a neural network based on the original training data, obtaining a second prediction value from the neural network based on the original training data using the trend information, determining a target loss by modifying the first loss based on the first prediction value and the second prediction value, and training the neural network based on the target loss.
Claims23
A neural network training method using trend, the neural network training method comprising:
- receiving original training data and trend information;
- obtaining a first prediction value and a first loss from a neural network based on the original training data;
- obtaining a second prediction value from the neural network based on the original training data using the trend information;
- determining a target loss by modifying the first loss based on the first prediction value and the second prediction value;
- and
- training the neural network based on the target loss.
A neural network training apparatus using trend, the neural network training apparatus comprising:
- a receiver configured to receive original training data and trend information;
- and
- a processor configured to obtain a first prediction value and a first loss from a neural network based on the original training data,
- obtain a second prediction value from the neural network based on the original training data using the trend information,
- determine a target loss by modifying the first loss based on the first prediction value and the second prediction value, and
- train the neural network based on the target loss.
A neural network training method using trend, the neural network training method comprising:
- receiving original training data and trend information;
- generating random training data based on a feature of the original training data;
- obtaining a first prediction value and a first loss from a neural network based on the random training data;
- adjusting the random training data using the trend information;
- obtaining a second prediction value from the neural network based on the adjusted random training data;
- determining a target loss by modifying the first loss based on the first prediction value and the second prediction value;
- and
- training the neural network based on the target loss.
Dependencies are inferred from the claim text.