norm => feature scaling

[c0ncentus]

https://en.wikipedia.org/wiki/Normalization_(statistics)

norm is "reScaling" or something like that

[wolfram77]

@c0ncentus Could you elaborate further?

[c0ncentus]

• for normalization you enter 3 inputs.

prefer 2 inputs like (value:number, to:[number, number])
more intuitive than "r" and "R"

[c0ncentus]

@c0ncentus Could you elaborate further?

"Norm" is too blurred for the usage its why I link wikipedia for telling you my purpose...

https://en.wikipedia.org/wiki/Normalized_number
nothing linked with this function named "norm"

[wolfram77]

@c0ncentus I see you are talking about API design. Do you thing the function name is confusing, or the arguments, or both? Currently the design is similar to Processing's norm.

[c0ncentus]

I already know,
normalization can say this things:

• system international (measure or unit) friendly (kilometers, or just kilo and so on)
• 2/10 => 1/5 (normalization)
• 200 => 10² * 2
• has 4 application on statistics
• ect ...

If you want to keep names because already written like this, it's up to you.
If I linked wikipedia is for telling you that's a wrong names that don't suit that purpose.

consequence of this : confusing, not sweet at all, dev should make an api over this api.

[wolfram77]

Wolfram has a function Rescale(x, {min, max}) and Rescale(x, {min, max}, {ymin, ymax}) (REF). R has rescale(x, to, from, ...) (REF), while norm(x, type = c("O", "I", "F", "M", "2")) refers to norm of a matrix (REF). Similar norm(A,p) with MATLAB (REF). MATLAB also has rescale(X,a,b) (REF). Processing has map(value, start1, stop1, start2, stop2) (REF). Python's sklearn has MinMaxScaler() (called rescaling) (REF). C++ has norm (const complex<T>& z) (REF).

We currently have remap(), which from the above reading suggests that rescale() is a better name. As for normalize(), we could call it minMaxScale() or minMaxNormalize(). For the time being, it's probably best to leave the old names as synonyms. meanNormalize() and standardize() may be reserved later. Thanks @c0ncentus for the suggestion and references.

References

• Normalization (statistics)
• Feature scaling
• How to normalize data to 0-1 range?