From the above we can see that firstly
a) Changing the euclidean distance to tangent distance accuracy increases as cluster becomes more dense or 
we can say that more distinguishable.
b) From the above plots it can be said that 1 and 7 are more distinguishable than 4 and 9 as mixing of cluster 
is more in later case.
c) From the plot for all digits for euclidean distance we can divide the numbers into sets(with most matching characteristics) are
 {0}, {{6,3,5}, {5,2,3},{8,5,1,2}}, {4,7,9} 
d)From the plot for all digits in tangent distance we can divide the numbers into sets(with most matching characteristics) are
 {0}, {2}, {4}, {7,9}, {6}, {5,3,8}, {1,2,8,5}