Do we need to reinitialized the weight using "init(net)" to reinitialized the weights in ANN training loop?

Hi, i need help on whether should re-initialized the weight when looping the training for different neurons number? will the training function re-initialized the weight by defautlt? the code i used are as follows. thank you very much.

N = [5 10 15 20 25 30 35 40 45 50 55 60];

trainFcn = 'trainlm'; % Levenberg-Marquardt backpropagation.

    %name the ANN NNxxpts
     NNVFBRReal200_5_to_60_all = cell(length(N),1) ;
     P = zeros(length(N),1) ;
for i = 1:length(N)
      % Create a Fitting Network & set number of neurons
      hiddenLayerSize = N(i); 
      net = init(net);
      net = fitnet(hiddenLayerSize,trainFcn);
      net.performFcn = 'mse';  % Mean Squared Error
% Choose Plot Functions
net.plotFcns = {'plotperform','plottrainstate','ploterrhist', ...
    'plotregression'};
      [net,tr] = train(net,X,T);
      nntraintool
      NNVFBRReal200_5_to_60_all{i} = net ;
      testX = X(:,tr.testInd);
      testT = T(:,tr.testInd);
      testY = net(testX);
      perf = mse(net,testT,testY) ;
      P(i) = perf ;

end as i encountered very high MSE when using trainBr. as below.

please help. however , in another case, it only occured at neurons = 35

Please help. thanks