Hi, I am reading an intereting paper titled 'Machine Learning in Network Centrality Measures'.
I have a Question: what is exactly the NN neuron composed of during your training?
Understanding: Paper mentions its Degree and Eigenvector centralities of synthetic network along with meta-parameters as input to NN and output label is Closeness and Betweenness centralities. (correct?)
- I asume you applied one time test execution for single whole synthetic network. (correct?)
- You proposed 20 neuron and three hidden layers is better model to train NN. please give me insight what the neuron composed of.
Thanks in advance.
Basharat Hussain
NN neuron is a new feature variable in the hidden layer. It is the weighted some of the variables in the input layer and the bias term.
You can see them as polinomial approximators with nonlinear functions. Therefore, the weights represents the relative importance of each input in its corresponding output.
Activation function is also playing role in transforming the resulting variable.
- Badges
- Science topic