Post Your Answer

In the Artificial neural network, what is the importance of activation function(s)?

3 years ago in Research Methods , Research Writing By Krirthi

Analysis Methodology Research methods

My research area is into wind forecasting and I was trying to implement a modified version of the Artificial neural network as a forecaster. I was trying to figure out the possible ways to implement ANN with changes over the activation function but struck up with errors. Could anyone specify to me the need for the activation function and what exactly it will do in ANN?

All Answers (2 Answers In All) Post Your Answer

By Sonam Bhatia Answered 3 years ago

As everyone is aware, neural networks consist of various layers like input, hidden and output. When an input is fed to the ANN, hidden nodes will accept only a linear combination of inputs. This linear output will be fed as input to the activation function where the complete operation takes place. If activation function is not employed, the results will be a simple linear combination of input which is of no use. Hence we formulate the activation function in order to provide non-linear space as well. Hence you can try out with an optimal combination of functionalities that can improve the ANN accuracy

Reply To Sonam Bhatia

By Sonali Jain Answered 3 years ago

Activation function is the core of the ANN network. The complete calculation/ processing of forecasting (as in your application) depends on how the activation function is deployed. For various input combinations (features), the activation function combines these linear inputs and based on the weight factor, resultant output is fed to the output layer. For improvisation in ANN, you may try using an optimal selection of weight factor instead of random weight selection.

Reply To Sonali Jain