20.1 Consider the problem of learning from objects' silhouettes (Figure 20.9). Calculate the entropy of the whole example set with respect to class, the residual information for attributes 'size' and 'holes', the corresponding information gains and gain ratios. Estimate the probabilities needed in the calculations simply by the relative frequencies, e.g. p(nut) = 3/12 or p(nut | holes-1) = 3/5.