%% Sample Data:
[x,t]=simplefit_dataset;

%% Initialize, train the 1-10-1 network:
net=feedforwardnet(10);
net=train(net,x,t);

%% Get some new data, then manually apply the network to the data:
xx=linspace(min(x),max(x),100);

% By default, Matlab scales the data.  We are using the parameters from the
% neural network and applying the same scaling to the new incoming data:
xa=mapminmax('apply',xx,net.input.processSettings{1});

% Here are the two sets of weights and biases:
W1=net.IW{1,1}; b1=net.b{1};
W2=net.LW{2,1}; b2=net.b{2};

% First prestate:
P1=W1*xa+repmat(b1,1,100);
% First state:
S1=tansig(P1);

% Second prestate (output layer, so state=prestate here)
P2=W2*S1+b2;

Y=mapminmax('reverse',P2,net.outputs{2}.processSettings{1});

% Plot:
plot(x,t,'.',xx,Y,'r-')