Research And Application Of Label Learning Based On Mixture Kernel Extreme Learning Machine

Because of the development of machine learning,a lot of machine learning algorithms have been proposed,among which Artificial Neural Networks(ANN)is one of the key research directions.The single hidden layer neural network is the most widely developed type of algorithm in the ANN.Traditional neural network algorithms such as the BP Propagation Neural Network have complex parameters,slow training,and huge data requirements,while the Extreme Learning Machine(ELM)is correct.The extension of the traditional single hidden layer neural network algorithm has greatly avoided the defects of the traditional neural network algorithm,so the introduction of the extreme learning machine into the problem of processing using the traditional neural network algorithm can greatly improve the performance.At the same time,with the increase of network usage,a large number of Internet users are generating huge amounts of data every day,and the data is very fragmented.The information density and usage value of these data are very low,so this requires a suitable method.To deal with these data,mark learning is a well-performing data classification method.It deals with the problem of mark recognition of unknown data by establishing a mapping of known features to marks.At present,there have been a lot of research results.Kernel function as an efficient dimensional space mapping method has important applications in artificial neural network algorithms,and kernel-based extreme learning machine is also an effective method to improve the performance of extreme learning machines.In this paper,a mixture kernel extreme learning machine is proposed for multi-label problem and gender label recognition.Multi-label learning algorithm based on kernel extreme learning machine can effectively improve the performance of mark classification,but most of the existing algorithms use a single kernel function,which cannot effectively solve the problem of data difference in multi-label.In this paper,mixture kernels are introduced into the extreme learning machine algorithm,and multi-label learning based on mixture kernel extreme learning machine is proposed.Firstly,the feature is mapped to the high-dimensional space by the mixed kernel function in the extreme learning machine algorithm.Then,the mixed kernel extreme learning machine model is used to obtain the output weight of the original label space.Finally,the model is used to predict the label of the unknown sample.Compared with the existing algorithms under identical data sets,the results show that the proposed algorithm outperforms multiple comparison algorithms under five evaluation indicators.The statistical hypothesis test further demonstrates the effectiveness of the proposed algorithm and is beneficial to multi-label learning algorithms.Performance improvement.At present,most of the gender marker recognition problems are based on traditional neural network algorithms.In this paper,a mixed kernel extreme learning machine with better performance in multi-marker recognition is applied to the gender labeling problem.The facial image features are extracted by LBP(Local Binary Pattern,LBP)and LPQ(Local Phase Quantization,LPQ),and the gender-marking is used to train the mixed kernel extreme learning machine to obtain the prediction model.The face image of the unknown gender marker is then extracted from the feature and the gender is determined by the prediction model.By comparing the recognition rates of other algorithms on the same public dataset,the effectiveness of the mixed kernel extreme learning machine in the gender labeling problem is proved.