6. (a) A group of botanists have collected data about the growth rate of trees (measured by the total mass of the tree at 10 years old), and related information about weather conditions and soil properties. They would like to use this to predict the growth rate of future trees. They have asked you to advise on whether they should use: (i) linear regression; (ii) logistic regression; or (iii) polynomial regression. In addition, provide an explanation on what kind of task each of these algorithms is suitable for, what kind of data each one needs. [7] (b) Describe in detail the batch gradient descent algorithm, as applied to multiple linear regression, including how to initialise values, details of how they are updated, and how to check for convergence. As part of your answer, provide a definition of the empirical error cost function used in this algorithm. [10] (c) Describe in detail an algorithm to build a linear classifier with multiple inputs, including an explanation of the perceptron learning rule. [8]