Study On The Change Of Texture Of Patinopecten Yessoensis Based On Artificial Neural Network

Patinopecten yessoensis is produced in the northerm coastal areas in China,it is famous for its nutrient and delicious taste.Texture is one of the characters that determines the quality of seafood,which is affected a lot by thermal processing,so in this essay,we built a prediction model by BP neural network and hope we can provide supporting for the storage and intensive processing of Patinopecten yessoensis.In this essay,we analyzed the texture data we got.We used MATLAB2008a to build a BP neural network,and predicted the texture parameters we got before.After programmed,tested and modified,a BP neural network with architecture of 2-3-1 is established,sigmoid activation function(tansig)for the input and hidden layer and the linear activation functions(purelin)for the output layer was selected,furthermore,the optimum parameters,training step of 5000,training goal of 0.001,are determined in order to provide high fault tolerance,reduce the convergence time,enhance the precision,and finally achieve accurate predictive results.All of the relative deviation of the parameters predicted with the model we built were under 10%,moreover,after being analyzed by zero mean deviation,X was all under 111.44,so the model was of highly credibility and fitting.