Aiming at training the feed-forward threshold neural network consisting of nondifferentiable activation functions, the approach of noise injection forms a stochastic resonance based threshold network that can be optim...Aiming at training the feed-forward threshold neural network consisting of nondifferentiable activation functions, the approach of noise injection forms a stochastic resonance based threshold network that can be optimized by various gradientbased optimizers. The introduction of injected noise extends the noise level into the parameter space of the designed threshold network, but leads to a highly non-convex optimization landscape of the loss function. Thus, the hyperparameter on-line learning procedure with respective to network weights and noise levels becomes of challenge. It is shown that the Adam optimizer, as an adaptive variant of stochastic gradient descent, manifests its superior learning ability in training the stochastic resonance based threshold network effectively. Experimental results demonstrate the significant improvement of performance of the designed threshold network trained by the Adam optimizer for function approximation and image classification.展开更多
Due to the demand of data processing for polar ice radar in our laboratory, a Curvelet Thresholding Neural Network (TNN) noise reduction method is proposed, and a new threshold function with infinite-order continuous ...Due to the demand of data processing for polar ice radar in our laboratory, a Curvelet Thresholding Neural Network (TNN) noise reduction method is proposed, and a new threshold function with infinite-order continuous derivative is constructed. The method is based on TNN model. In the learning process of TNN, the gradient descent method is adopted to solve the adaptive optimal thresholds of different scales and directions in Curvelet domain, and to achieve an optimal mean square error performance. In this paper, the specific implementation steps are presented, and the superiority of this method is verified by simulation. Finally, the proposed method is used to process the ice radar data obtained during the 28th Chinese National Antarctic Research Expedition in the region of Zhongshan Station, Antarctica. Experimental results show that the proposed method can reduce the noise effectively, while preserving the edge of the ice layers.展开更多
The innovations actually diffuse among social network nowadays.Individual heterogeneity,interactions between individuals and network topology influence a lot.We established a "double threshold" modified mode...The innovations actually diffuse among social network nowadays.Individual heterogeneity,interactions between individuals and network topology influence a lot.We established a "double threshold" modified model and took the number of neighbors,neighbors' adoption and the cost-benefit parameters as crucial influencing factors.The diffusion of DaLingTong(CDMA450)products in MeiShan city of SiChuan province during 2004 to 2007 has been used to verity the model on Matlab.The validation results fit the actual diffusion pattern of DaLingTong(CDMA450) products very well.The results indicate that there exists a "tipping point(threshold)" in the process of innovation diffusion.If the initial adoption quantity is larger than the tipping point,then the product will spread to a large portion of people,otherwise is will collapse to zero.The model can effectively predict the diffusion of new products,and can influence the diffusion process by changing the value of the parameters.展开更多
To improve the accuracy of modulated signal recognition in variable environments and reduce the impact of factors such as lack of prior knowledge on recognition results,researchers have gradually adopted deep learning...To improve the accuracy of modulated signal recognition in variable environments and reduce the impact of factors such as lack of prior knowledge on recognition results,researchers have gradually adopted deep learning techniques to replace traditional modulated signal processing techniques.To address the problem of low recognition accuracy of the modulated signal at low signal-to-noise ratios,we have designed a novel modulation recognition network of multi-scale analysis with deep threshold noise elimination to recognize the actually collected modulated signals under a symmetric cross-entropy function of label smoothing.The network consists of a denoising encoder with deep adaptive threshold learning and a decoder with multi-scale feature fusion.The two modules are skip-connected to work together to improve the robustness of the overall network.Experimental results show that this method has better recognition accuracy at low signal-to-noise ratios than previous methods.The network demonstrates a flexible self-learning capability for different noise thresholds and the effectiveness of the designed feature fusion module in multi-scale feature acquisition for various modulation types.展开更多
基金Project supported by the Natural Science Foundation of Shandong Province,China(Grant No.ZR2021MF051)。
文摘Aiming at training the feed-forward threshold neural network consisting of nondifferentiable activation functions, the approach of noise injection forms a stochastic resonance based threshold network that can be optimized by various gradientbased optimizers. The introduction of injected noise extends the noise level into the parameter space of the designed threshold network, but leads to a highly non-convex optimization landscape of the loss function. Thus, the hyperparameter on-line learning procedure with respective to network weights and noise levels becomes of challenge. It is shown that the Adam optimizer, as an adaptive variant of stochastic gradient descent, manifests its superior learning ability in training the stochastic resonance based threshold network effectively. Experimental results demonstrate the significant improvement of performance of the designed threshold network trained by the Adam optimizer for function approximation and image classification.
基金Supported by the National High Technology Research and Development Program of China (No. 2011AA040202)the National Natural Science Foundation of China (No. 40976114)
文摘Due to the demand of data processing for polar ice radar in our laboratory, a Curvelet Thresholding Neural Network (TNN) noise reduction method is proposed, and a new threshold function with infinite-order continuous derivative is constructed. The method is based on TNN model. In the learning process of TNN, the gradient descent method is adopted to solve the adaptive optimal thresholds of different scales and directions in Curvelet domain, and to achieve an optimal mean square error performance. In this paper, the specific implementation steps are presented, and the superiority of this method is verified by simulation. Finally, the proposed method is used to process the ice radar data obtained during the 28th Chinese National Antarctic Research Expedition in the region of Zhongshan Station, Antarctica. Experimental results show that the proposed method can reduce the noise effectively, while preserving the edge of the ice layers.
基金ACKNOWLEDGEMENTS Project supported by the National Social Science Foundation of China （Grant No. 11BGL041）, Ministry of Education Humanities and Social Sciences General Project （12YJA630166）.
文摘The innovations actually diffuse among social network nowadays.Individual heterogeneity,interactions between individuals and network topology influence a lot.We established a "double threshold" modified model and took the number of neighbors,neighbors' adoption and the cost-benefit parameters as crucial influencing factors.The diffusion of DaLingTong(CDMA450)products in MeiShan city of SiChuan province during 2004 to 2007 has been used to verity the model on Matlab.The validation results fit the actual diffusion pattern of DaLingTong(CDMA450) products very well.The results indicate that there exists a "tipping point(threshold)" in the process of innovation diffusion.If the initial adoption quantity is larger than the tipping point,then the product will spread to a large portion of people,otherwise is will collapse to zero.The model can effectively predict the diffusion of new products,and can influence the diffusion process by changing the value of the parameters.
基金Project supported by the National Key R&D Program of China(No.2020YFF01015000ZL)the Fundamental Research Funds for the Central Universities,China(No.3072022CF0806)。
文摘To improve the accuracy of modulated signal recognition in variable environments and reduce the impact of factors such as lack of prior knowledge on recognition results,researchers have gradually adopted deep learning techniques to replace traditional modulated signal processing techniques.To address the problem of low recognition accuracy of the modulated signal at low signal-to-noise ratios,we have designed a novel modulation recognition network of multi-scale analysis with deep threshold noise elimination to recognize the actually collected modulated signals under a symmetric cross-entropy function of label smoothing.The network consists of a denoising encoder with deep adaptive threshold learning and a decoder with multi-scale feature fusion.The two modules are skip-connected to work together to improve the robustness of the overall network.Experimental results show that this method has better recognition accuracy at low signal-to-noise ratios than previous methods.The network demonstrates a flexible self-learning capability for different noise thresholds and the effectiveness of the designed feature fusion module in multi-scale feature acquisition for various modulation types.
