Training method and device of attribute prediction model, attribute prediction method, device and equipment
By first training a pre-trained model of chemical substance structure labels and then adjusting it using a small amount of attribute label data, the problem of insufficient data in compound/protein attribute prediction models is solved, and the prediction accuracy is improved.
CN114758729BActive Publication Date: 2026-06-19TENCENT TECHNOLOGY (SHENZHEN) CO LTD +1
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- TENCENT TECHNOLOGY (SHENZHEN) CO LTD
- Filing Date
- 2021-01-11
- Publication Date
- 2026-06-19
Smart Images
Figure CN114758729B_ABST
Abstract
This application discloses a training method, apparatus, and device for an attribute prediction model, relating to the field of chemical molecular attribute prediction. The method includes: acquiring a first set of sample data, comprising a first chemical substance and its structural labels, the structural labels describing the atomic structure of the chemical substance; training a pre-trained model based on the first set of sample data, the pre-trained model outputting the structural labels of the input chemical substance; acquiring a second set of sample data, comprising a second chemical substance and its attribute labels, the attribute labels describing the properties of the chemical substance; and training an attribute prediction model based on the second set of sample data and the pre-trained model, the attribute prediction model outputting the attribute labels of the input chemical substance. This method can reduce the amount of labeled data required to train the attribute prediction model.
Need to check novelty before this filing date? Find Prior Art