Neural Network Model Construction Methods Based On Multi-View Learning And Its Related Applications

With the development of artificial intelligence-related technologies,multi-view learning has been applied and practiced in many fields.From the continuous enrichment of multi-view data acquisition scenes to the emergence of novel multi-view collaborative learning modeling methods,multi-view learning has shown good application potential in every link.Compared with traditional single-view learning,multi-view learning can not only ensure the independence of data from each view,but also obtain the correlation information between views,thus improving the generalization ability of the model.In this paper,the modeling process of multiview learning is explored and practiced,and the focus of the work is on the collaborative learning method of view,and two kinds of multi-view collaborative learning classification models are proposed.(1)MV-RBF,a dual-view collaborative learning classification model for time series data.Based on the radial basis neural network model,this method can analyze and apply the internal information related to the data from multiple views by introducing cooperative learning items with the ability of cooperative learning from views.The feasibility and effectiveness of the model are verified by experiments on multi-view data set composed of multiple stock data.(2)Multi-view collaborative learning classification model for three or more views.In order to improve the flexibility of the model,a view weighting mechanism is introduced to give corresponding weight to the characteristic information provided by each view data.Then,more cooperative learning between views is realized by adding the cooperative learning ability item of views to the objective function of the model.The introduction of view weighting mechanism and collaborative learning mechanism can ensure the independence of view and capture more relevant information between views.In the experimental part,the feasibility of the new multiview collaborative learning classification model is verified by conducting classification experiments on multiple sets of stock data and EEG data,and analyzing the stability of the constructed model and the factors that affect the experimental results.