Identifying the most influential features (factors, input or variable) or variable importance called in other modeling techiques is in interesting topic, and certainly neuron killing topic too.

There has been puzzling me for night and day, a lot research papers on it, and here is a very good summary of it.

And the easiest way to do it? Not twisting brains to figure out the computer brain. Do feature selection prior running the model, or else running decision tree or random forest based on the scored results with nn predicted result as target then identify the most influential features.


 (_,=-   )
        `} Ojo