JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD
JAWAHARLAL NEHRU TECHNOLOGICAL UNIVERSITY HYDERABAD MCA 2019 MACHINE LEARNING Previous Question Paper (December - 2019 - R17)
Note: This question paper contains two parts A and B. Part A is compulsory which carries 25 marks. Answer all questions in Part A. Part B consists of 5 Units. Answer any one full question from each unit. Each question carries 10 marks and may have a, b, c as sub questions.
PART - A
List and explain the issues in machine learning.
Answer
Machine learning faces several fundamental challenges: Data Quality and Quantity, where models require large, clean, and representative datasets to avoid bias or overfitting; Overfitting and Underfitting, which occur when a model either captures noise as patterns or fails to learn the underlying structure; Computational Complexity, involving the high cost of training deep models or processing high-dimensional data; Interpretability, often referred to as the 'black box' problem where complex models lack transparency in decision-making; and Generalization, the difficulty of ensuring a model performs accurately on unseen, real-world data.
Discuss in brief about neural network representation.
Answer
Neural network representation is based on a directed graph of interconnected processing units called neurons, organized into layers. The Input Layer receives raw data, Hidden Layers perform non-linear transformations using weights and activation functions, and the Output Layer produces the final prediction. Each connection has an associated weight that determines the strength of the signal, and each neuron typically applies a non-linear activation function, such as Sigmoid or ReLU, allowing the network to approximate complex, non-linear functions.
Describe briefly about Bayesian Belief Networks.
Answer
Sign in to view the complete answer paper.
What are temporal patterns? Explain.
Answer
Sign in to view the complete answer paper.
Briefly discuss about explanation based learning.
Answer
Sign in to view the complete answer paper.
PART - B
In the following, it is desired to describe whether a person is ill. We use a representation based on conjunctive constraints (three per subject) to describe individual person. These constraints are “running nose”, “coughing”, and “reddened skin”, each of which can take the value true (+) or false (-). We say that somebody is ill, if he is coughing and has a running nose. Each single symptom individually does not mean that the person is ill. Specify the space of hypotheses that is being managed by the version space approach. Arrange all hypotheses in a graph structure using the more-specific-than relation.
Answer
Sign in to view the complete answer paper.
Table below shows the relationship between the body height and the gender of a group of persons (the records have been sorted with respect to the value of height in cm). Calculate the information gain for potential splitting thresholds and determine the best one.
| Height | 161 | 164 | 169 | 175 | 176 | 179 | 180 | 184 | 185 |
|---|---|---|---|---|---|---|---|---|---|
| Gender | F | F | M | M | F | F | M | M | F |
Answer
Sign in to view the complete answer paper.
Describe how the basic Back-Propagation Learning Algorithm is used in Multi-Layer Network.
Answer
Sign in to view the complete answer paper.
Consider a learned hypothesis, h, for some boolean concept. When h is tested on a set of 100 examples, it classifies 83 correctly. What is the standard deviation and the 95% confidence interval for the true error rate for ErrorD(h)?
Answer
Sign in to view the complete answer paper.
Describe in detail about Maximum Likelihood Hypotheses for Predicting probabilities.
Answer
Sign in to view the complete answer paper.
Give an example to explain the concept of K-Nearest neighbor algorithm.
Answer
Sign in to view the complete answer paper.
Explain in detail about Dynamic Time Warping Methods.
Answer
Sign in to view the complete answer paper.
Explain how training and testing is performed in discrete hidden markov models.
Answer
Sign in to view the complete answer paper.
Describe how prior knowledge is used to alter the Search Objective.
Answer
Sign in to view the complete answer paper.
Illustrate how learning can be performed using inductive-analytical approach.
Answer
Sign in to view the complete answer paper.