Joe designed two neural networks: NN1 and NN2. Both networks have a softmax output layer with three classes and a cross-entropy loss. He feeds an input from class 1 to each network. The output of NN1 is 0.41, 0.37, and 0.22 for classes 1, 2, and 3, respectively. The output of NN2 is 0.74, 0.11, and 0.15 for classes 1, 2, and 3, respectively. Calculate the loss of each network and report it for NN1 and NN2 Based on these findings, (NN1 or NN2) is a better design.

Joe designed two neural networks: NN1 and NN2. Both networks have a softmax output layer with three classes and a cross-entropy loss. He feeds an input from class 1 to each network. The output of NN1 is 0.41, 0.37, and 0.22 for classes 1, 2, and 3, respectively. The output of NN2 is 0.74, 0.11, and 0.15 for classes 1, 2, and 3, respectively. Calculate the loss of each network and report it for NN1 and NN2 Based on these findings, (NN1 or NN2) is a better design.

Database System Concepts

7th Edition

ISBN:9780078022159

Author:Abraham Silberschatz Professor, Henry F. Korth, S. Sudarshan

Publisher:Abraham Silberschatz Professor, Henry F. Korth, S. Sudarshan

Chapter1: Introduction

Section: Chapter Questions

Problem 1PE

See similar textbooks

Similar questions

Write a python program based on the discussion we had on the last day of class, to train a network with two hidden layers each with 5 neurons. Finish the gradient computation we did in class, and implement a steepest decent search for a network which minimizes the cost function. Once you believe that your network is trained, compute the value of n(1, 1) and see what you get. Bonus: If you easily get this to work, then try to generalize to K neurons in each of the two hidden layers. Does the training work better with more neurons?
w1 w5 i1 h1 01 w2 w6 w3 12 02 h2 w4 b1 Figure 1: Neural Network You are provided with a simple NN as shown in Figure 1. The inputs, target output and initial weights of the network are as below: il- .os i2 ol-.01 -10 02=.99 bl-.35 b2= .60 w1-.15 w2-.20 w3=.25 w4-.30 w5=.40 w6-.45 w7=.50 w8=.55 O Find the updated weights of w8 and w4 after the first iteration of your NN training. Use the leaning rate for gradient descent as 0.5 Round all calculation to nine decimal digits.
Write the MATLAB code for the following
Construct a Turing machine that takes strings over the alphabet {a,b} on the tape, and moves the whole string over to the right one cell position. When you're done, the tape should essentially look the same (since it's infinite on both ends…) but you will know (and people who look at your code will know) that in fact you moved the string over by one space to the right. After moving the string over, the turing machine should move to the left-most cell in the string and than go to the HALT state.
Q4/ The structure of a neural network consists of [2:3:2]. It is trained using a BPA. The weights and biases of a hidden layer are: Wh = 3 While the weights and biases of the output layer are: bh= - 1²/31 Wo= --[7²2²2²31₁ b. - [1³3] = Determine the MSE at the first attrition when the desired output about [0,1], if the input values are [0.9, 0.1).
Draw a recurrent network with ‘4’ neurons, but with no feedback from the same neuron by including the unit delay in output? (please the solution, enter it written on the keyboard)
3.
Construct a Turing machine that takes strings over the alphabet {a,b} on the tape, and moves the whole string over to the right one cell position. When you're done, the tape should essentially look the same (since it's infinite on both ends…) but you will know (and people who look at your code will know) that in fact you moved the string over by one space to the right. After moving the string over, the turing machine should move to the left-most cell in the string and than go to the HALT state.
A perceptron has three inputs X₁, X₂, X3 and a bias input fixed to 1. It classifies using a threshold step function [h(z)= 1 if z>0; else 0] as its activation function. Initialize the corresponding weights as follows: w₁ = 0.7, W₂ = 0.2, W3 = 0.9. Let the learning rate be 0.2 and bias weight w. be -0.7. W₁ X1 W₂ X₂ 0 W3 WA X3 +1 Apply the perceptron learning rule with the above input and with desired output 0. What are the new weights after this update ? Note: The output 0 here is fixed and may or may not be the answer to the previous question.
Suppose we have four machines that are using the Berkeley algorithm for clock synchronization. The time daemon is running on machine MO. The other machines are named M1, M2, and M3. The time readings on the four machines are as follows wenty-four hour clocks are usedt. MO 09:10 M1: 08:30 M2 10:10 M3: 07:14 In the third step of the algorithm, what are the values that the time daemon sends to each machine? Assume that the time values are expressed in minutes. Below, these values are presented in a vector value sent to MO, value sent to MI, value sent to M2. value sent to M3I none of these O84, 44, 96, 321 O(24, 16, -84, 92) O192. 84, 16. -24 O-32, 76, 40
P9.4 The operational amplifier-resistor neuron circuit from Figure P9.4(a) needs to be analyzed. Compute fo2 as a function of w₁, W₂, and w3, and of the bias T as indicated on the neuron symbol of Figure 9.4(b). Assume for computation purposes that the two neurons are equivalent. 600 W 30002 W 180002 9000 ww 12 k ww Www 18000 H 3 Xxz X4 =-4V X₁ X₂ Xy W₁ W₂ for (a) 12000 www for -f(x-7) W3 T -1 (b) Figure P9.4 High-gain neuron with weights and nonzero threshold: (a) circuit diagram and (b) symbol. (a) Draw the neuron's transfer characteristics as a function of z= Wix. (b) Determine the range of activation z that allows for the linear operation of the neuron. Assume fat-| = |fsat+= 14 V. (c) Assume that the neuron needs to classify patterns of two classes and find its responses due to the eight individual input values, which are vertices of a three-dimensional [-1,1] cube.
Define a fully connected network with two hidden layers with Nh1 and Nh2 features, respectively. The hidden layers should use relu activation and the final layer should not have any activation. a linear network with no hidden layer. We will use this network to study the benefit of depth, or equivalently using a non-linear network instead of a linear network. The final layer should not have any activation class NeuralNet(N.Module): # YOUR CODE HERE class LinearNet(N.Module): #YOUR CODE HERE