Property model training method and apparatus, electronic device, computer-readable storage medium, and computer program product

An electronic device, comprising:

• a memory, configured to store executable instructions;
• and
• a processor, configured to execute the executable instructions stored in the memory to implement a molecular property model training method including:
• obtaining a plurality of pieces of first sample data associated with a plurality of chemical reactions, each first sample data comprising a plurality of pieces of first molecule data related to a corresponding chemical reaction and indicator data of the chemical reaction;
• performing forward propagation on the plurality of pieces of first molecule data of the plurality of pieces of first sample data in a first encoding network to obtain a first encoding feature of each first molecule data;
• determining an uncertainty loss of the plurality of chemical reactions based on the first encoding features of the plurality of pieces of first molecule data of the plurality of pieces of first sample data and indicator data of each chemical reaction, the uncertainty loss being used for fitting uncertainty of establishment of the plurality of chemical reactions;
• and
• performing backward propagation processing on the uncertainty loss in the first encoding network to obtain an updated first encoding network as a molecular property model, the molecular property model being used for determining a property of a target molecule.

A non-transitory computer-readable storage medium, having computer executable instructions stored thereon, the computer executable instructions, when executed by a processor of an electronic device, causing the electronic device to implement a molecular property model training method including:

• obtaining a plurality of pieces of first sample data associated with a plurality of chemical reactions, each first sample data comprising a plurality of pieces of first molecule data related to a corresponding chemical reaction and indicator data of the chemical reaction;
• performing forward propagation on the plurality of pieces of first molecule data of the plurality of pieces of first sample data in a first encoding network to obtain a first encoding feature of each first molecule data;
• determining an uncertainty loss of the plurality of chemical reactions based on the first encoding features of the plurality of pieces of first molecule data of the plurality of pieces of first sample data and indicator data of each chemical reaction, the uncertainty loss being used for fitting uncertainty of establishment of the plurality of chemical reactions;
• and
• performing backward propagation processing on the uncertainty loss in the first encoding network to obtain an updated first encoding network as a molecular property model, the molecular property model being used for determining a property of a target molecule.