Fundamentals of Machine Learning — Winter 2021

Define following terms with respect to machine learning:

precision (ii) overfitting (iii) concept learning

What are the issues in machine learning as far as designing of learning system is concerned?

Asample dataset about stolen cars is given below: Sample No. Color Type Origin Stolen? 1 Red Sports Domestic Yes 2 Yellow Sports Domestic No 3 Red Sports Imported Yes 4 White Sports Imported No 5 White Family Imported Yes 6 White SUV Domestic No 7 Yellow SUV Imported Yes 8 Yellow Family Domestic No 9 Red SUV Imported No 10 Red Sports Domestic Yes Find the probability of a stolen car which is white in color, sports type and origin domestic. Use naïve bayes classification.

Define following terms with respect to machine learning:

hypothesis (ii) hypothesis space (iii) hypothesis space search

Briefly explain the steps to prepare the version space for given training dataset.

Is given statement true or false? Reduced error pruning is following pre-pruning approach. Explain one-by-one step to perform reduced error pruning to get optimized decision tree.

With suitable example, show one-by-one step of backpropagation algorithm on multilayer feedforward artificial neural network.

Define following terms with respect to machine learning:

target function (ii) inductive learning hypothesis (iii) version space

Explain inductive bias with suitable example.

What is semi-supervised learning? Explain the steps of self-training algorithm. OR1

Define following terms with respect to machine learning:

consistent hypothesis (ii) underfitting (iii) mean square error

Compare supervised learning with unsupervised learning.

What is perceptron? What is perceptron training rule? What are its limitations and by which rule these limitations can be overcome?

For which type of problem characteristics, decision tree algorithm is suitable?

Define following terms with respect to machine learning:

gradient descent (ii) learning rate (iii) cost function (iv) epoch

Asample dataset about diagnostic is given below: Sample Fever Vomiting Diarrhea Shivering Diagnosis No. 1 no no no no healthy 2 average no no no influenza 3 high no no yes influenza 4 high yes yes no salmonella poisoning 5 average no yes no salmonella poisoning 6 no yes yes no bowel inflammation 7 average yes yes no bowel inflammation Using gain ratio, find the purest attribute that can be at root of the decision tree.

Define: (i) prior probability (ii) posterior probability (iii) maximum likelihood

State Bayes theorem and its applications.

What is EM algorithm? Explain it in detail.

What is Bayesian Belief Network?

Briefly explain minimum description length principle.

Define: precision, recall. Make confusion matrix from given data and find precision, recall, accuracy and F- score from it: No. of samples in test data having class value POSITIVE = 150 No. of samples in test data having class value NEGATIVE = 600 No. of POSITIVE samples of test data classified as POSITIVE by model = 100 No. of NEGATIVE samples of test data classified as NEGATIVE by model = 550

Briefly explain Bayes optimal classifier.

What is recommender system? How machine learning methods can be useful in it?

Define: sensitivity, specificity. Make confusion matrix from given data and find sensitivity, specificity, accuracy and F-score from it: No. of samples in test data having class value RED = 250 No. of samples in test data having class value GREEN = 450 No. of RED samples of test data classified as RED by model = 200 No. of GREEN samples of test data classified as GREEN by model = 400