There is no built-in function in Matlab to round a number to specific significant figures. However I have found a little trick from somewhere in the internet which works! It simply makes use of the 'mat2str' (or 'num2str') and 'eval' (or 'str2num') Matlab functions. For example, first generate a matrix 'A' with random numbers:

A = rand(2,3)*100

A =
70.6046 27.6923 9.7132
3.1833 4.6171 82.3458

Then convert into 'string' using 'mat2str' and specifies how many digits of precision, here I have made this to be 3:

>> A_temp = mat2str(A,3)

A_temp =
[70.6 27.7 9.71;3.18 4.62 82.3]

Now convert the string back to numerical values by using 'eval':

>> A_rounded = eval(A_temp)

A_rounded =
70.6000 27.7000 9.7100
3.1800 4.6200 82.3000

So the values have been rounded to 3 sig. fig. :-)

Now if we use 'num2str' instead of 'mat2str', we need to convert back to numerical values by using 'str2num' because 'eval' will not work. Moreover, I just found out that using 'num2str' can actually allow rounding to specific decimal places. Using the same 'A' matrix as above, and then specify the precision to be '%4.4e\n' which will round the values to the 4th decimal place:

A_temp2 = num2str(A,'%4.4e\n')

A_temp2 =
7.0605e+001
2.7692e+001
9.7132e+000
3.1833e+000
4.6171e+000
8.2346e+001

>> A_rounded2 = str2num(A_temp2)

A_rounded2 =
70.6050
27.6920
9.7132
3.1833
4.6171
82.3460

Noticed that it is necessary to have the '\n' in the string that specifies the format of precision, otherwise the result will go weird.

PS: I noticed that there are still small rounding errors (or floating point errors), however this trick should be good enough for general purposes.