Q4(a) Give two examples of how machine learning impacts the applications you use regularly. How well do you think machine learning works in those instances?

Q4(b). Machine learning already works very well on facial recognition. Give one example of where this could be beneficial, and one example where this might be used to do harm. How would you suggest that potential misuse could be moderated?

Q4(c) Explain why it is important in machine learning to use training data which are: (i) large, (ii) good-quality, and (iii) representative.

Q4(d) What are common techniques used to detect outliers? Explain DBSCAN algorithm? What are the benefits and downsides of removing outlisers from data? How does presence of outliers impact variance in statistical learning?

Q4(e) The following set of two-dimensional objects should be clustered:

Object feature1 feature2

O1 1 4

O2 2 1

O3 3 7

O4 4 3

O5 5 6

O6 8 5

O7 9 3

O8 9 7

O9 10 6

O10 11 2

with the following scatter diagram:

8 7 5 feature 2 4 3 2 O 2 6 CO 10 12 feature 1\f