Sun, Y., Zhang, M., Bhandari, B., & Yang, P. (2019). Intelligent detection of flavor changes in ginger during microwave vacuum drying based on LF-NMR. Food Research International, 119, 417-425.
Sun, Y., Zhang, M., Mujumdar, A. S., & Yu, D. (2020). Pulse-spouted microwave freeze drying of raspberry: Control of moisture using ANN model aided by LF-NMR. Journal of Food Engineering, 110354.
30. The author need to explain this part in more details. These lines are not clear. Does the 3-10-1 describes the topology of the hidden layer or the topology of the neural network as a whole? What was the reason to use the Levenberg-Marquardt algorithm? Isn’t this algorithm implemented by the trainlm function in matlab?
[Response] Thanks for your comments. The 3-10-1 describes the topology of the neural network as a whole, the neuron number of input layer, hidden layer, and output layer was 3, 10, and 1, respectively. The trainlm is the Levenberg-Marquardt backpropagation. Algorithm is the basis of neural network technology, which directly determines the generalization ability of network model, nonlinear mapping ability, fitting precision and efficiency, etc. Levenberg-Marquardt algorithm is the fastest algorithm proposed for medium-sized networks. In the pre-experiment, we studied the effects of different algorithms on the model of French fries by training time, MSE and R of training, validation, testing, and overall data. The results showed that the best performing algorithm was Levenberg-Marquardt.
31. How did the author reach this conclusion? Is there any reference that says an R2>0.863 is an indicator of good prediction ability?! In addition, a testing set should be generated using a new set of oil and potatoes. My understanding is that the author divided the data that was collected in the same experiment to 3 sets. That is not how external validation is conducted.
[Response] Thanks for your comments. We have revised the description in the Manuscript. The results showed that the R2 of the validation set of TPC, TGP, THP, AV, PV were >0.930 and >0.940, respectively, indicating that the prediction ability of the model was available (table 1) ,numerous studies revealed that the R2 higher than 0.9 in the model indicated excellent classification performance as well as prediction ability (Cavanna, Righetti, Elliott, & Suman, 2018). However, R2 of the testing set of OTG was only 0.863, which is not good owing to the fluctuation of OTG with frying time. In addition, we have added the external validation in the Manuscript. Line 239-245
32.Again, what tools or model indicators did the author use to reach this conclusion?