I am working on machine learning data were i get different length from each sample. For example i get 40 columns from first sample then i get 45 for next sample and for some samples i get more.
i want to resale (a kind of normalization) all sample to specific number of columns such that i don't loss the actual distribution of data.
Statistics, machine learning, Neural Networks.
Hello, you need to make interpolation function, which is built-in option in many contemporary mathematical packages and codes. Using interpolation function you can variate the needed scale and number of colums. But it it needed to be accurate, because the interpolation can be done by various ways, for example linear or spline type. What is better depends on the preliminary information and intuitive suggestion.
Yes Samer,
I want to make same features for all of the instances as it is now different. How could be it is possible that feature don't disturb its distribution.
http://www.cs.xu.edu/math/math116/12f04/Ch06.pdf
https://www.stt.msu.edu/academics/classpages/uploads/SS15/200-4/Chapter6.pdf
- Badges
- Science topic