人工神经网络试题及答案..docVIP

Question One: The weight updating rules of the perceptron and Kohonen neural network are _____. Question Two: The limitation of the perceptron is that it can only model linearly separable classes. The decision boundary of RBF is __________linear______________________whereas the decision boundary of FFNN is __________________non-linear___________________________. Question Three: The activation function of the neuron of the Perceptron, BP network and RBF network are respectively________________; ________________; ______________. Question Four: Please present the idea, objective function of the BP neural networks (FFNN) and the learning rule of the neuron at the output layer of FFNN. You are encouraged to write down the process to produce the learning rule. Question Five: Please describe the similarity and difference between Hopfield NN and Boltzmann machine. 相同： Both of them are single-layer inter-connection NNs. They both have symmetric weight matrix whose diagonal elements are zeroes. 不同： The number of the neurons of Hopfield NN is the same as the number of the dimension (K) of the vector data. On the other hand, Boltzmann machine will have K+L neurons. There are L hidden neurons Boltzmann machine has K neurons that serves as both input neurons and output neurons (Auto-association Boltzmann machine). Question Six: Please explain the terms in the above equation in detail. Please describe the weight updating equations of each node in the following FFNN using the BP learning algorithm. (PPT原题y=φ(net)= φ(w0+w1x1+w2x2)) W0=w0+ W1=w1+ W2=w2+ Question Seven: Please try your best to present the characteristics of RBF NN. (1)RBF networks have one single hidden layer. (2)In RBF the neuron model of the hidden neurons is different from the one of the output nodes. (3)The hidden layer of RBF is non-linear, the output layer of RBF is linear. (4)The argument of activation function of each hidden neuron in a RBF NN computes the Euclidean distance between input