An algorithm is said to learn successfully from data and experience E for a task T, if the algorithm is satisfying a performance measure P. Considering following tasks, what would you recommend E and P for each of these tasks:

(a) Automated fraud detection in bank transactions

(b) Automated hand-writing detection

Q3.

Which type of learning supervised, unsupervised and reinforcement learning is suitable for the following scenarios, please justify your answer:

(a) UPS wants to group the regions of higher activity based on the orders received weekly from different regions so that UPS can deploy more resources in higher activity regions. UPS doesnt have any labeled or marked data to dig out this information, so the UPS IT team needs to think of an appropriate solution in this case.

(b) A big real estate firm wants to predict property prices in the future based on past data, market influences, and economic policies affecting the purchasing power of customers. Which type of learning algorithm is best suited for this firm?

(c) Optimizing chemical reactions for different experimental conditions such as temperature, pressure, pH, etc. by varying these conditions and awarding credit or loss points with regards to the resulting product of the chemical reactions which type of learning can provide promising results here?

Q4. Consider following four vectors: 4 u = (6, 5), v = (6, -5), w = (-5, 6), x = (-5,-6) Which pairs of vectors are at right angles to each other? How?

Q5. Find eigenvalues and eigenvectors of matrix:

1	4
1	1

A = 1 1 4 1