Hi Nazila,
Given a 2-D matrix of size 700x9, where the task is to predict a price, you could use the following code structure to define a BiLSTM network:
% Defining the layers of the BiLSTM network
layers = [
% numFeatures = 9 in your case as input is of dimension 700x9
sequenceInputLayer(numFeatures)
% 100 is the number of hidden units in bilstm
bilstmLayer(100, 'OutputMode', 'sequence')
% '1' specifies that the fully connected layer has only one neuron used
% to output a single continuous value
fullyConnectedLayer(1)
];
net = trainnet(XTrain,YTrain,layers,"mse",options);
While defining the layers, you have the option to use a "regressionLayer", but it is not recommended to use it now according to the documentation. Instead, you can use the "trainnet" function and specify the loss function as "mse" within the training options.
Also make sure your input is in the format of sequences. To know more refer this documentation: https://www.mathworks.com/help/deeplearning/ref/trainnet.html#mw_8b6c966a-f840-4426-9693-279222b8372fv
You can refer to the documentation links below for more information:
Hope this helps!!
