Dear Peinan Wu

Once you get your desire ANN model, you have to generate a function by clicking 'MATLAB Function' (Fig. 2).

Then you have to save the function (default name is 'myNeuralNetworkFunction').

(keep file name and function same) (Fig. 4)

Again, you have to create a custom function using 'myNeuralNetworkFunction'

For instance, the function name 'funct' (Fig. 5) (keep file name and function name same).

Now, your objective function is 'funct'. Here, a negative sign is given because the present study is for maximization. (GA toolbox only do minimization, therefore the function has been modified with negative sign). For minimization problem, please keep it positive.

Again, this function can be used for response prediction.

Apart from GA, you can use PSO code for optimization. (Link for code: https://yarpiz.com/50/ypea102-particle-swarm-optimization )

Do not worry, there will be a negative value of an optimum objective function after GA optimization. However, you have to take the mod value of the function.

Recently I have published one article on the application of ANN, SVM, GA and PSO in a biological system (using MATLAB). This paper might help you.

Article Optimization of dark fermentative hydrogen production from o...

You can follow these link to create ANN model in MATLAB

https://www.youtube.com/watch?v=ZORch7uFJ5s

https://www.youtube.com/watch?v=8edTpED2abw

Let me know if any other help is required

[email protected]

Thank you

As far as your question is concerned it is a regression task. You can code a simple algorithm in python using tensorflow rather than resorting to inbuilt matlab functions as python allows for more flexibility. But If it is an optimization task it is better that you use grasshopper, rhino3d ... etc for youre purpose.

Feel free to contact if your doubt persist.

Dear Peinan Wu

Once you get your desire ANN model, you have to generate a function by clicking 'MATLAB Function' (Fig. 2).

Then you have to save the function (default name is 'myNeuralNetworkFunction').

(keep file name and function same) (Fig. 4)

Again, you have to create a custom function using 'myNeuralNetworkFunction'

For instance, the function name 'funct' (Fig. 5) (keep file name and function name same).

Now, your objective function is 'funct'. Here, a negative sign is given because the present study is for maximization. (GA toolbox only do minimization, therefore the function has been modified with negative sign). For minimization problem, please keep it positive.

Again, this function can be used for response prediction.

Apart from GA, you can use PSO code for optimization. (Link for code: https://yarpiz.com/50/ypea102-particle-swarm-optimization )

Do not worry, there will be a negative value of an optimum objective function after GA optimization. However, you have to take the mod value of the function.

Recently I have published one article on the application of ANN, SVM, GA and PSO in a biological system (using MATLAB). This paper might help you.

Article Optimization of dark fermentative hydrogen production from o...

You can follow these link to create ANN model in MATLAB

https://www.youtube.com/watch?v=ZORch7uFJ5s

https://www.youtube.com/watch?v=8edTpED2abw

Let me know if any other help is required

[email protected]

Thank you

You can go through my publication, it also makes of NN MATLAB toolbox

Article A Differential Evolution Based Adaptive Neural Network Pitch...

Chandan Mahata gave a detailed answer. Apart from GUI, you can also use the command prompt. After loading the data in your workspace (remember the rows in the data are your independent factors/predictors and columns are the runs/experiments/observations), create a neural network using the command

net = newff(input, target, n); where 'input' is the predictors, 'target' is wind speed and n is the number of nodes in the input layer.

You can set other parameters like performance functions, epoch etc with

net.performFcn = 'mae';

net.trainParam.epochs=5000;

(See help for more details)

To train the network use command

net = train(net, input, target );

For prediction use command

Trained = sim(net, value); where 'value' is the condition on which you want to predict the wind speed.

You should divide your data into two parts use one part to train the network and the other part to check the performance of the network.

To check the performance of the network use

plotregression(target, predicted) 'predicted' is the wind speed predicted by the trained network