Stopping a Myth in Artificial Neural Networks

摘要

The myth of "the more neurons, the better learning" is commonly held in artificial intelligence personnel. The side effect of having too many neurons may not be easily observed unless some active measurements and analyses are taken. In this study, I performed a series of experiments to make various neural networks learn the same quadratic equation and then analyzed three statistical data to illustrate the problems that can be caused by having more than necessary number of neurons in a neural network.