Neural Network Approximation: Three Hidden Layers Are Enough

A three-hidden-layer neural network with super approximation power is introduced. This network is built with the floor function (), the exponential function (), the step function (), or their compositions as the activation function in each neuron and hence we call such networks as Floor-Exponential-Step (FLES) networks. For any width hyper-parameter , it is shown that FLES networks with width and three hidden layers can uniformly approximate a H\"older continuous function on with an exponential approximation rate , where and are the H\"older order and constant, respectively. More generally for an arbitrary continuous function on with a modulus of continuity , the constructive approximation rate is . Moreover, we extend such a result to general bounded continuous functions on a bounded set . As a consequence, this new class of networks overcomes the curse of dimensionality in approximation power when the variation of as is moderate (e.g., for H\"older continuous functions), since the major term to be concerned in our approximation rate is essentially times a function of independent of within the modulus of continuity. Finally, we extend our analysis to derive similar approximation results in the -norm for via replacing Floor-Exponential-Step activation functions by continuous activation functions.
View on arXiv