why my training results are repeated even after i run my code again and again ??

% ANN model for ukraine with all inputs

%% Data loading to the matlab environment

train_data = xlsread('Train_M.xlsx');

input_train = xlsread('Train_M','input');

X_train = input_train';

target_train = xlsread('Train_M','target');

T_train = target_train';

test_data = xlsread('Test_M.xlsx');

input_test = xlsread('Test_M','input');

X_test = input_test';

target_test = xlsread('Test_M','target');

T_test = target_test';

%% normalization of training and testing data

%for training

[Tr_i ps] = mapminmax(X_train,-1,1);

[Tr_t ps] = mapminmax(T_train,-1,1);

%for testing

[Tst_i pp] = mapminmax(X_test,-1,1);

[Tst_t pp] = mapminmax(T_test,-1,1);

%clear

%clc

%load Best_Dataset (need to save our datadivision for avoiding

%randomization)

%% GRNN model development

%% model formulation

spread = 0.42;

net = newgrnn(Tr_i,Tr_t,spread);

view(net);

%% prediction outputs of the model

y_tr = net(Tr_i);

y_tst = net(Tst_i);

%% renormalization process

% Model outputs(forecasted data)

Renorm_train_out = mapminmax('reverse',y_tr,ps);

Renorm_test_out = mapminmax('reverse',y_tst,pp);

%for model inputs(observed data)

Renorm_train_inp = mapminmax('reverse',Tr_t,ps);

Renorm_test_inp = mapminmax('reverse',Tst_t,pp);

%% performance of the ANN model

%Training performance and testing performance

plot(Renorm_test_inp,Renorm_test_out,'x')%curves is ok

plot(Renorm_train_inp,Renorm_train_out,'x')%curve is ok

%% statistical measure of the model

%% model performance command

mdl = fitlm(Renorm_test_inp,Renorm_test_out)%testing performance

mdl = fitlm(Renorm_train_inp,Renorm_train_out)%training performance

%% saving the trained model

save('GRNN_allinputs.mat','net');

%% the results for the above code repeates.

R2 is 0.68, RMSE 10.8 - training

R2 is 0.639 ,RMSE 3.03 -testing

these results are again shown for the second run and so on

AM i missing something here ???

please help !!!!

Answer 1

[Himanshu]

Hi,

I see that you are facing repeated training results even after running your code multiple times.

The reason behind this is likely due to the deterministic nature of the Generalized Regression Neural Network (GRNN). GRNN does not have random initialization, so it will always produce the same results given the same inputs. Ensure that data normalization is consistent and correctly applied to avoid any discrepancies.

Please refer to the below documentations for more information.

1. Generalized Regression Neural Networks: https://www.mathworks.com/help/deeplearning/ug/generalized-regression-neural-networks.html
2. Design generalized regression neural network: https://www.mathworks.com/help/deeplearning/ref/newgrnn.html
3. Process matrices by mapping row minimum and maximum values to [-1 1]: https://www.mathworks.com/help/deeplearning/ref/mapminmax.html

I hope this helps.