Abstract: There are proposed methods, devices, and computer program products for extending a feature space of a data sample. In the method, a global representation is obtained for a feature in a plurality of features of the data sample. A local representation is obtained for the feature based on a classifying criterion for classifying the data sample into one of a plurality of predefined domains. A representation is generated for the feature of the data sample based on the global representation and the local representation. With these implementations, an exclusive feature space may be created for each domain identified by the classifying criterion, which is dedicated to capturing domain-specific knowledge and characteristics.

Abstract: There are proposed methods, devices, and computer program products for feature management. In the method, a first event associated with a first and a second object, and a second event associated with the first and second events are obtained, and a type of the first event is different from a type of the second event. A first feature of the first object is determined based on a first encoder, and a second feature of the second object is determined based on a second encoder. The first encoder is updated based on the first and second features and the first and second events. With these implementations, multiple events are used in determining the encoder for extracting the feature, and thus the encoder may have better performance in accuracy and increase performance of downstream tasks.

Abstract: An embodiment of the present disclosure discloses a virtual reality (VR) rehabilitation training method with a quantitative evaluation function. The method includes: obtaining attribute information of a user to be trained, and playing short scenarios based on the attribute information; during the playing, acquiring pupil change information and action information of the user in real time, so as to determine a training scenario that the user is interested in and a training solution; generating a virtual training scenario based on the training scenario that the user is interested in; obtaining a head position of the user to be trained, and generating a virtual body model; generating a virtual distractor in the virtual training scenario based on the training solution; and tracking and recording somatosensory signals of the user to be trained at six degrees of freedom, and performing functional quantitative analysis to obtain functional quantitative results.

Abstract: There are proposed methods, devices, and computer program products for prediction model management. In the method, gradient information associated with the prediction model is obtained based on sample data for a time slot in a predetermined time period. An offset of the time slot in the predetermined time period is acquired. A step size is determined for updating a parameter of the prediction model based on the gradient information, the offset, and historical gradient information that is determined based on historical sample data for a group of historical time slots before the time slot. With these implementations, the whole training procedure may be divided into multiple time period and each time period may further include multiple time slots. During each time period, the offset may be used to control the importance of the historical gradient information and the gradient information in determining the step size.