Redirigiendo al acceso original de articulo en 18 segundos...
Inicio  /  Algorithms  /  Vol: 14 Par: 2 (2021)  /  Artículo
ARTÍCULO
TITULO

Effects of Nonlinearity and Network Architecture on the Performance of Supervised Neural Networks

Nalinda Kulathunga    
Nishath Rajiv Ranasinghe    
Daniel Vrinceanu    
Zackary Kinsman    
Lei Huang and Yunjiao Wang    

Resumen

The nonlinearity of activation functions used in deep learning models is crucial for the success of predictive models. Several simple nonlinear functions, including Rectified Linear Unit (ReLU) and Leaky-ReLU (L-ReLU) are commonly used in neural networks to impose the nonlinearity. In practice, these functions remarkably enhance the model accuracy. However, there is limited insight into the effects of nonlinearity in neural networks on their performance. Here, we investigate the performance of neural network models as a function of nonlinearity using ReLU and L-ReLU activation functions in the context of different model architectures and data domains. We use entropy as a measurement of the randomness, to quantify the effects of nonlinearity in different architecture shapes on the performance of neural networks. We show that the ReLU nonliearity is a better choice for activation function mostly when the network has sufficient number of parameters. However, we found that the image classification models with transfer learning seem to perform well with L-ReLU in fully connected layers. We show that the entropy of hidden layer outputs in neural networks can fairly represent the fluctuations in information loss as a function of nonlinearity. Furthermore, we investigate the entropy profile of shallow neural networks as a way of representing their hidden layer dynamics.

 Artículos similares

       
 
Ismaila A. Oyehan, Ajiboye S. Osunleke and Olanrewaju O. Ajani    
The dynamics of a quadruple tank system (QTS) represent an extensive class of multivariate nonlinear uncertain systems found in the industry. It has been established that changes in split fractions affect the transmission zero location, thereby altering ... ver más
Revista: ChemEngineering

 
Jiaqi Xue, Biao Ma, Man Chen, Liang Yu and Liangjie Zheng    
This paper investigates the effects of disc deformation on the tilting-induced vibration of a splined spinning disc with axial-fixed boundaries.The purpose is to provide an intuitive interpretation of the vibration variance of the wet clutch system with ... ver más
Revista: Applied Sciences

 
Yi Hu, Yong Zhao and Haopeng Liang    
This paper proposes a novel fully nonlinear refined beam element for pre-twisted structures undergoing large deformation and finite untwisting. The present model is constructed in the twisted basis to account for the effects of geometrical nonlinearity a... ver más
Revista: Aerospace

 
Yifan Lu, Qi Shao, Liangliang Lv, Guangqiang Fang and Honghao Yue    
Due to the super flexibility and strong nonlinearity of space membrane antennas, the dynamic response of a space membrane antenna will be affected by the rigid?flexible coupling effect in the process of orbital maneuvering. In this case, the dynamic mode... ver más
Revista: Aerospace

 
Dominik Schäfer    
The numerical assessment of T-tail flutter requires a nonlinear description of the structural deformations when the unsteady aerodynamic forces comprise terms from lifting surface roll motion. For linear flutter, a linear deformation description of the v... ver más
Revista: Aerospace