In artificial neural network, I am using back-propagation to optimize the matrix weight. Can I do further improvement of my network by adjusting individual values of steepness for each neuron. I am using sigmoid function as the activation function.
Thanks in advance,
Nothing prevents you from doing it. Also, when you know the function you are trying to approximate has certain characteristics, you can even choose a different activation function than a sigmoid, which will approximate the function faster, leading to a faster training process.
However, for the steepness, check this article: http://www.pjoes.com/pdf/25.4/Pol.J.Environ.Stud.Vol.25.No.4.1467-1477.pdf
You can read the following:
https://in.mathworks.com/help/nnet/ug/optimize-neural-network-training-speed-and-memory.html
- Badges
- Science topic
- Similar topics
- Mathematics
- Algebra
- Matrix