Scribd
Upload
28 views3 pages

CS462 Assignment2

This document contains a neural networks assignment with multiple choice questions covering topics like perceptrons, activation functions, learning modes, CNNs, SVMs and more. It also includes discussion questions about stochastic vs batch learning, cross-validation, components of CNNs, advantages of CNNs, kernel tricks and Cover's theorem.

Uploaded by

Irene Gabriel
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
28 views3 pages

CS462 Assignment2

This document contains a neural networks assignment with multiple choice questions covering topics like perceptrons, activation functions, learning modes, CNNs, SVMs and more. It also includes discussion questions about stochastic vs batch learning, cross-validation, components of CNNs, advantages of CNNs, kernel tricks and Cover's theorem.

Uploaded by

Irene Gabriel
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Misr University for Science and Technology

College of Information Technology

CS462: Neural Networks, Fall 22,


by: Prof. Dr. Mostafa Gadalhaqq
Assignment 2
A. Multiple Choice Questions
 Modeling the XOR logic gate by the perceptron was a challenge because:
a. It is hard to model c. It represent a nonlinearly seprable data
b. It represent a linearly seprable data d. None of the above
 If the input vector {4, 10, 5, 20} is applied to a 4-input neuron with weights 4, 3, 2 and 1 and the
activation function is f(v)=2v. If the bias is equal to 0, the output will be:
a. 76 a. 123
b. 238 b. None of the above
 Which is true for neural networks?
a. It has set of nodes and connections c. Node could be in excited state or non-excited state
b. Each node computes it’s weighted input d. All of the mentioned
 One can solve a non-linearly separable data with
a. The preceptron c. A multilayer neural networks
b. A single layer neural network d. All of the above
 Which is false for neural networks?
a. It has set of nodes and connections c. Nodes can be fully or partially connected
b. MLP can have any number of hidden layers d. Activation of hidden nodes can be step function
 Which of the following is an activation function?
a. ReLU function c. Swish function
b. Sigmoid function d. All of the above function
 The activation function that is now mostly used for MLP
a. ReLU function c. tanh function
b. Sigmoid function d. (b) or (c)
 The activation function that is now mostly used for RNN
a. ReLU function c. tanh function
b. Sigmoid function d. (b) or (c)
 For a binary classification problem, which of the following activation function is used?
a. ReLU fuction c. Softplus fuction
b. Sigmoid fuction d. Binary step fuction
 Stochastic learning mode is computationally faster than the batch learning mode
a. true
b. false
 In stochastic learning mode network weights are updated for each sample
a. true
b. false

1
Misr University for Science and Technology
College of Information Technology

 In ............. , adjustment of the weights of the network is performed after the epoch
a. Stochastic mode c. Machine Learning mode
b. Batch mode d. Adaptive Learning mode
 In ............. , adjustment of the weights of the network is performed on sample-by-sample basis
a. Stochastic mode c. Machine Learning mode
b. Batch mode d. Adaptive Learning mode
 CNN is the acronyms of
a. Concurrent Neural Networks c. Convolutional Neural Networks
b. Convectional Neural Networks d. All of the above
 Which stage in the pattern recognition system the Convolution layers play its role?
a. Data Acqquisition c. Feature Extraction
b. Preprocessing d. Classification
 The RBF is a multilayer network?
a. true
b. false
 In the SVM, support vectors are chosen data samples that:
a. Have non-zero alpha value c. Maximize the margin between classes
b. Have zero alpha value d. (a) and (c)

B. Discuss briefly the following terminologies


C. Stochastic Learning mode
• Adjustment of the weights of the MLP are performed on an example-by-example basis.
Guarantee for convergence to local minimum. Well suited for non linear regression problems
D. Batch Learning mode
• Adjustment of the weights of the MLP is performed after the presentation of all the N training
examples. It is less likely to be trapper in a local minimum. .
E. Cross-validation
• Cross-Validation is a standard tool in statistics that provide appealing guiding principle where
the training set is used to assess various models and choose the “best” one
F. CNN
• A class of artificial neural networks that designed to automatically learn spatial features
through backpropagation by using multiple building blocks, such as convolution layers, pooling
layers, and fully connected layers.
G. RBF
• Input layer: Consists of n source nodes (n is the dimensionality of x)
• Hidden Layer: Consists of the same number of computational units as the size of the training
samples.
• Output layer: Consists of a single (or more) computational unit
H. SVM
 Given a training samples, the SVM constructs a hyperplane as decision surface in such a way
2
Misr University for Science and Technology
College of Information Technology
the margin of separation between positive and negative examples is maximized.

I. Compare between the CNN and the MLP.

 Both MLP and CNN can be used for Image classification however MLP takes vector as
input and CNN takes tensor as input so CNN can understand spatial relation(relation
between nearby pixels of image)between pixels of images better thus for complicated
images CNN will perform better than MLP
J. Discuss the different components of the CNN and list some of the famous CNN
architectures.
 Convolution is a mathematical operation on an image and kernel to detect a
particular feature from the input image and get the result having information
about that feature
 Pooling is the process of reducing the size of the data while keeping the most
important features
 Flattening is converting the data into a 1-dimensional array for inputting it to the
next layer.
K. Discuss the advantages and disadvantages of the CNN.
 (Adv) Very accurate at image recognition and classification.
 (Adv) Weight sharing
 (Adv) CNNs make use of the same knowledge across all image locations.
 (DisAdv) A lot of training data is needed for the CNN to be effective.
 (DisAdv) CNNs tend to be much slower because of operations like maxpool.
L. Discuss the kernel trick and state the main condition that should be satisfied in
the kernel.
 The “trick” is that kernel methods represent the data only through a set of
pairwise similarity comparisons between the original data observations x (with
the original coordinates in the lower dimensional space), instead of explicitly
applying the transformations ϕ(x)‫ز‬
M. State and Discuss Cover’s Theorem, and prove its correctness with a numeric
example.
 A dichotomy {C1, C2} is said to be φ-separable if there exist a m1-dimensional
vector w such that we may write

You might also like