Abstract: Embodiments of the disclosed technologies include generating a reward score for an entity. A rate distribution is determined using the reward score. A sampled rate value is generated by sampling the rate distribution. A probability score is generated for a pair of the entity and a user using the sampled rate value. A probability distribution is determined using the probability score. A sampled probability value is generated by sampling the probability distribution. A machine learning model is trained using the sampled probability value.

Abstract: Embodiments of the disclosed technologies include generating a reward score for an entity. A rate distribution is determined using the reward score and a number of times the entity has been selected for ranking. A sampled rate value is generated by sampling the rate distribution. A probability score is generated for a pair of the entity and a user based on the sampled rate value. A probability distribution is determined using the probability score. A sampled probability value is generated by sampling the probability distribution. A machine learning model is trained using the sampled probability value.

Abstract: Embodiments of the present invention disclose a method for processing cascade conference sites in a cascade conference, which may be used to implement one-to-one correspondence between an image orientation and a voice orientation of each conference site in the cascade conference, and improve the user experience of a participant. Certain embodiments of the present invention include: receiving an audio code stream sent by a cascade conference site, where the audio code stream sent by the cascade conference site is sent to different conference sites through different audio sound channels or audio cascade channels; receiving an audio code stream sent by a non-cascade conference site; and selecting audio data satisfying a preconfigured condition from audio data to be selected.

Abstract: A method, a device and a system for mixing processing of an audio signal are provided in the embodiments of the present invention. The method includes: judging a channel type of a receiving terminal; for a single-channel receiving terminal, sending a mixed audio signal and meanwhile sending location information of a sending terminal that has maximum audio signal energy on each sub-band of the mixed audio signal to the single-channel receiving terminal; for a double-channel receiving terminal or a multi-channel receiving terminal, performing up-mixing to obtain double-channel or multi-channel audio data according to location information that is allocated to a single-channel sending terminal, performing mixing processing on audio data that participates in mixing to obtain double-channel or multi-channel mixed audio data, and sending the double-channel or multi-channel mixed audio data.