2-
Using your favorite Neural Net toolbox or MATLAB code for patternnet, develop a
model using any learning rate from [1e-2 1e-1] with your desired hidden layer neurons.
Note, you cannot modify learning rate through toolbox, hence default rate is fine.
a. Run the model and comment on its performance. Open up confusion matrix and
comment on the results.
Hint: if using MATLAB code, create a neural network, configure with (nameneural
network name, inputs, targets), add learning rate as training parameter, assign a
performing function ‘mse’, divide dataset to train, validate and test ratio, train
network
b. Run the model with multiple hidden layers and learning rates (possible only through
MATLAB code at this time). Comment on its performance.
Pass number of hidden layers as parameter when creating neural network
Ex. numHidden = ([20,15,10]); net = patternnet(numHidden)
c. Is there anything else that can be done to the model to improve performance