神经网络英语演讲.pptxVIP

神经网络英语演讲第1页/共15页 TitleA review on neural networks with random weights.第2页/共15页 Contents1 Shallow feed-forword neural network with random weights(NNRWs).2 Deep neural network with random weights.3 Conclusion.第3页/共15页 第4页/共15页 第5页/共15页 1 Shallow feed-forword neural network with random weights 1.1 TuringThe father of AI(artificial intelligence)Turing TestCan Machines Think?第6页/共15页 1.2 perception Retina of sensory unitsA set of association cellsResponse units第7页/共15页 1.3 RVFL(random vector functional link networks)[-1,1]MooreCPenrose pseudo-inverse.第8页/共15页 1.4 ELM(extreme learning machine) ZeroOne unified formgeneralized第9页/共15页 2 Deep NNRWs(neural networkwith random weights)2.1 AE(auto-encoder)+NNRWs3-layerUnsupervised learning第10页/共15页 2.2 RBM(restricted boltzmann machines)+NNRWs Two layersFully connectedNo connections in the same layer第11页/共15页 3 conclusionAdvantages:Faster learning speedImprove the computing efficiency.Disadvantages:The randomization range. The type of distribution of the hidden weights.第12页/共15页 Cao W, Wang X, Ming Z, et al. A Review on Neural Networks with Random Weights[J]. Neurocomputing, 2017.第13页/共15页 Thank you第14页/共15页 感谢观看！第15页/共15页