Hello
I have questions
In ANN at matlab software:
What is the meaning
1.gradiant 0.10341 in epoch 1000
2.validation check=0 at epoch 1000
3.mu=.0.01 at epoch 1000?
Thanks
Epoch is the itration
Validation check : https://in.mathworks.com/matlabcentral/answers/351812-how-do-validation-check-work-in-neuralnet
m u. :
mu is the control parameter for the algorithm
for more see here :
https://m.youtube.com/watch?v=IHZwWFHWa-w
MatLab provides many toolboxes for modelling and simulations including ANNs, Deep Learning, and Optimizations.
Dear Mohammad,
Check through the matlab documentation of the toolbox.
Best regards
Epoch is the itration
Validation check : https://in.mathworks.com/matlabcentral/answers/351812-how-do-validation-check-work-in-neuralnet
m u. :
mu is the control parameter for the algorithm
for more see here :
https://m.youtube.com/watch?v=IHZwWFHWa-w
Dear Mohammad,
These values are related to the performance of the ANN after 1000 epochs such as gradient. Normally we use an epsilon with a very small value, if gradient return a value less than epsilon, then it means that ANN was implemented correctly. The other two are related to validation check = 0 at 1000 it means that no validation check is done.
The mu value is used to control the weights of the neurons updating process .
Dear Mohamad,
This depend on the problem at hand. In general when you create your ANN network in MATLAB you have what is called training parameters such as minimum gradient which is th epsilon in our case and by default it is a positive value of 1e-07. This means is the minimum performance gradient before training
is stopped. When the performance gradient becomes too small, continued
training is unlikely to produce significant improvements.
in your case it is not yet this value, but as I said it depends on your problem and how you set your minimum gradient.
Mu is a blending factor. The greater it is the more weight is given to
gradient descent learning and a small step size. The smaller it is the
more weight is given to large step sizes with Newton's method.
So depending on how you want to run your network you set your mu value.
Finally, for validation check see the following MATHWORKS answer to the same problem.
https://www.mathworks.com/matlabcentral/answers/276887-problem-with-validation-check-number-for-matlab-neural-network
Hello
Please send your viewpoint.
Can you help me as a co author for publishing paper in springer journal?
Hello
Can you help me in SOM neural network?
My draft will provide.
- Badges
- Science topic