Abstract: The present disclosure relates to a data protection method, apparatus, medium and electronic device. The method comprises: obtaining a specified batch of reference samples of an active participant of a joint training model; determining generation gradient information of the first reference sample; determining target gradient information sent to the passive participant according to the generation gradient information, and sending the target gradient information to the passive participant, to update, by the passive participant, parameters of the joint training model according to the target gradient information. Through the above solution, the influence of the generated data on the training process and model performance of the joint training model is avoided as much as possible, and the privacy and security of data are improved.

Abstract: The present disclosure describes techniques for determining performance of a classifier. A first machine learning model and a second machine learning model may be trained by aggregating updates to the first machine learning model and the second machine learning model received from a plurality of client computing devices. A cumulative distribution function (CDF) associated with a distribution of the positive samples in the user data may be estimated using the trained first machine learning model. A probability density function (PDF) associated with a distribution of the negative samples in the user data may be estimated using the trained second machine learning model. An integration-based computation of an area under the receiver operating characteristic curve (AUC) of the classifier may be performed using the PDF and the CDF.

Abstract: A method includes receiving an utterance of speech and segmenting the utterance of speech into a plurality of segments. For each segment of the utterance of speech, the method also includes extracting a speaker=discriminative embedding from the segment and predicting a probability distribution over possible speakers for the segment using a probabilistic generative model configured to receive the extracted speaker-discriminative embedding as a feature input. The probabilistic generative model trained on a corpus of training speech utterances each segmented into a plurality of training segments. Each training segment including a corresponding speaker-discriminative embedding and a corresponding speaker label. The method also includes assigning a speaker label to each segment of the utterance of speech based on the probability distribution over possible speakers for the corresponding segment.

Abstract: The present disclosure describes techniques for improving data subsampling for recommendation systems. A user-item graph associated with training data may be constructed. An importance of user-item interactions may be estimated via graph conductance based on the user-item graph. An importance of the training data may be measured via sample hardness using a pre-trained pilot model. A subsampling rate may be generated based on the importance estimated from the user-item graph and the importance measured by the pre-trained pilot model.

Abstract: A method includes receiving an utterance of speech and segmenting the utterance of speech into a plurality of segments. For each segment of the utterance of speech, the method also includes extracting a speaker=discriminative embedding from the segment and predicting a probability distribution over possible speakers for the segment using a probabilistic generative model configured to receive the extracted speaker-discriminative embedding as a feature input. The probabilistic generative model trained on a corpus of training speech utterances each segmented into a plurality of training segments. Each training segment including a corresponding speaker-discriminative embedding and a corresponding speaker label. The method also includes assigning a speaker label to each segment of the utterance of speech based on the probability distribution over possible speakers for the corresponding segment.

Abstract: A method includes receiving an utterance of speech and segmenting the utterance of speech into a plurality of segments. For each segment of the utterance of speech, the method also includes extracting a speaker-discriminative embedding from the segment and predicting a probability distribution over possible speakers for the segment using a probabilistic generative model configured to receive the extracted speaker-discriminative embedding as a feature input. The probabilistic generative model trained on a corpus of training speech utterances each segmented into a plurality of training segments. Each training segment including a corresponding speaker-discriminative embedding and a corresponding speaker label. The method also includes assigning a speaker label to each segment of the utterance of speech based on the probability distribution over possible speakers for the corresponding segment.

Abstract: A method includes receiving an utterance of speech and segmenting the utterance of speech into a plurality of segments. For each segment of the utterance of speech, the method also includes extracting a speaker=discriminative embedding from the segment and predicting a probability distribution over possible speakers for the segment using a probabilistic generative model configured to receive the extracted speaker-discriminative embedding as a feature input. The probabilistic generative model trained on a corpus of training speech utterances each segmented into a plurality of training segments. Each training segment including a corresponding speaker-discriminative embedding and a corresponding speaker label. The method also includes assigning a speaker label to each segment of the utterance of speech based on the probability distribution over possible speakers for the corresponding segment.