Scribd
Upload
23 views

Question Bank of Advanced Dbms

The question for adbms

Uploaded by

Swathi SB
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
23 views

Question Bank of Advanced Dbms

The question for adbms

Uploaded by

Swathi SB
Copyright
© © All Rights Reserved
Available Formats
Download as DOCX, PDF, TXT or read online on Scribd
You are on page 1/ 2

QUESTION BANK

ARTIFICIAL NEURAL NETWORKS

1. Explain how the example described by a vector of continuous attributes is forward-


propagated through the multilayer perceptron. How is the network’s output
interpreted?
2. What is the transfer function? Write down the formula defining the sigmoid transfer
function, and describe its shape.
3. Write down the formulas defining the error rate and the mean square error.
4. Why is the training technique called “backpropagation of error”? What is the rationale
behind the need to establish the “neurons’ responsibilities”?
5. Write the pseudo code for Backpropagation of error in a neural network with one
hidden layer.
6. What are the benefits of an adaptive learning rate? What formula has been
recommended for it?
7. What do you know about the danger that the training of a neural network might result
in overfitting the training data?
8. Explain the algorithm that searches for a reasonable size of the hidden layer. What
main difficulties does it face?

DECISION TREES

1. Describe the mechanism that uses a decision tree to classify examples.


2. Explain the principle of the divide-and-conquer technique for induction of decision
trees.
3. What are the advantages of small decision trees in comparison to larger ones?
4. What determines the size of the decision tree obtained by the divide-and-conquer
technique?
5. What is entropy and how does it relate to the frequency of the positive and negative
examples in the training set?
6. How do we use entropy when assessing the amount of information contributed by an
attribute?
7. What mathematical finding has reduced the number of thresholds that need to be
investigated?
8. Write the algorithm for decision-tree pruning.
9. What are the potential benefits of decision-tree pruning?
10. How can we estimate the tree’s error rate on future data? Write down the formula and
explain how it is used.
11. Explain the mechanism that converts a decision tree to a set of rules. How many rules
are thus obtained? What is the motivation behind such conversion?
Exercise problems

UNSUPERVISED LEARNING

1. What is the main difference between unsupervised learning and supervised learning?
2. What is a cluster of examples? Define some mechanisms to describe the clusters.
3. How can we measure the distance between an example and a cluster?
4. Summarize the main benefits of cluster analysis.
5. Write the pseudo code for clustering algorithm k-means.
6. What would be the consequence if we did not normalize the training set? Write down
the simple normalization formula.
7. Describe some methods for the initialization of k-means. What are the main
consequences of good or bad initialization?
8. Explain the principle of hierarchical application of k-means.
9. Write the basic algorithm of hierarchical aggregation.
10. What kind of clusters cannot be detected by the k-means algorithm?
11. What distance metric is used in hierarchical aggregation? What are the advantages
and disadvantages of this metric?
12. Describe the principle of the hierarchical-aggregation approach to clustering.

You might also like