Patent attributes
Methods, systems, apparatuses, and computer programs, are described for generalizing a learned behavior across different tasks. In one aspect, a method includes obtaining first data that describes sensed attributes of a first environmental state, obtaining second data that defines a target end state after performance of a particular task, obtaining first output data generated by an affective experience module that represents a particular behavior to be performed by an agent system to complete the particular task in the environment, providing, as an input data to a machine learning model that has been trained to generate second output data indicative of a particular behavior that can be used to complete the task in the environment based on processing, by the machine learning model, of the input data, the input data comprising the first data, the second data, and the first output data, obtaining the second output data generated by the machine learning model, and selecting a particular behavior for enactment to complete the particular task based on the second output data.
