How does a very small number behave while processing?

Question

Well I am working on a big dataset and after some calculations I am getting values for the features like 4.4E-5. I read it somewhere those values means 0.000044 that is ten to the power minus 5. So my question is whenever I want to use them for further processing will these values behave same as float works or do I need some other data type?

Answer 1

Yes, it is an extended notation presenting the same binary floating point data type. Both 4.4E-5 and 0.00044 are the same. And that value only approximates 0.000044 with a sum of powers of 2: 2^-18 + ...

Answer 2

Multiplying lots of small numbers leads to underflow. Take the log and add. This technique is universal in computer science. Many of the Google hits for "underflow log" are useful, including SO hits, other techniques for dealing with it, etc.