X1 ai X2 laz Figure 2 A Neural Network For Classification (a) For classification, compared with Naïve Bayes classifier,

[answerhappygod]

X1 Ai X2 Laz Figure 2 A Neural Network For Classification A For Classification Compared With Naive Bayes Classifier 1 (144.56 KiB) Viewed 27 times

X1 ai X2 laz Figure 2 A Neural Network For Classification (a) For classification, compared with Naïve Bayes classifier, explain a unique advantage of Neural Networks, and a unique disadvantage of Neural Networks. [6 marks] (b) The neural network architecture given in Figure 2 uses non-linear neurons with LeakyReLU activation function a(z) = { z.if z>0 }{0.12, if z 50. Let a be the activation of unit i at layer j (i.e., aº is the output of the unit i at layer j, after applying the activation function). Let o be the weight parameter on the link from a to a 0) C+1) The cost function is defined as J(0) = (a) – ya)*+(a) – ya)?. (Notations follow -Y2 our lecture). Consider a single training sample (32) = (0:2) and the corresponding ground-truth label () = (.). Use the training data to train the neural network in Figure 2, and solve it by gradient descent with learning rate 0.1, and find the values of all weight parameters oi, and the value of the cost function, after the first iteration. The initial settings of the weights are as follows. Please show calculation details. (Keep values in 3 decimal places) [18 marks] = Initial weights: 0.0 = 0.3 0.1 = 0.2 60 = 0.8 2 = 0.7 02 = -0.6 = -0.3 01 = -0.1 02 = 0.2 0,2 = 0.4 02 = 0.6 02 (1) 20 = = 0.2 = 0.4 0.3 = -0.5 30 32