ANN SCG train function
9 views (last 30 days)
Show older comments
Hi,
I have some questions about Matlab ANN algorithm functions.
1) I have an example about ANN which is used trainscg as the train function .What does train function means ? Because we know that we can add so many layers and every layer has own activation function. I think trainscg function is just used help to minimize error in loss function (like cross-entropy) , right ? So it is not activation function.I think it is just an optimizer . But why matlab called it train function ?
I want to replicate this function in Python but I have no idea which function I should use instead of trainscg. How about ADAM or SGD ?
2) Is there any difference between tangent sigmoid activation function and sigmoid function activation function?
Thanks.
0 Comments
Answers (1)
Vineet Joshi
on 26 Mar 2021
Hi
Network training functions are functions in MATLAB used to train a neural network using a specified algorithm.
trainscg for example uses scaled conjugate gradient backpropagation as a training algorithm where as traingdx uses gradient descent with momentum and adaptive learning rate backpropagation as a training algorithm.
Apart from a train function, a network in MATLAB uses other function properties like adaptation function or performance function.
The details of which can be found here.
As for your other question, they key difference between both is activation function is for tansig (Tangent Sigmoid) the range is between [-1 and 1] whereas for logsig (Sigmoid) the range is [0,1] although both make an S-shaped curve.
Hope this helps.
See Also
Categories
Find more on Sequence and Numeric Feature Data Workflows in Help Center and File Exchange
Community Treasure Hunt
Find the treasures in MATLAB Central and discover how the community can help you!
Start Hunting!