How do I add features to a fully connected layer in a MATLAB neural network?

I am running a LSTM network on some input data using the trainNetwork function. The data consists of sequencing data with nine features, and I have broken the sequences into windowed segments that I can classify. I want to pass the average, standard deviation, and other features of these segments to the fully connected layer of my neural network to hopefully improve the accuracy of the classifier. I have tried using an addition layer to add the statistical features and features from the LSTM output together but I have had no success. Is this even possible in MATLAB, and if so, how would one go across implementing this?

Answer 1

[Srivardhan Gadila]

You can write your own Custom Layer and place it before the fullyConnectedLayer in order to achieve the above funcitonality. To get started you can refer to Define Custom Deep Learning Layers & Deep Learning Custom Layers. Also the following answer may help you in the process: Can anyone help me in reshaping a fully connected layer output to a image?