Feature processing: A little experience

I. feature vector Normalization

First, we need to clarify that the normalization of feature vectors is essentially different from that of feature vectors. The feature vector normalization is entrywise, that is, for each vector element. The feature vector standardization is to transform the vector to a space with a "length" of 1.

1. linear function conversion, the expression is as follows:

Y = (X-minvalue)/(maxvalue-minvalue)

2. logarithm function conversion, the expression is as follows:

Y = log10 (X)

3. The expression for Inverse cotangent Function Conversion is as follows:

Y = arctan (x) * 2/PI

4. Subtract the mean value and multiply the variance:

Y = (X-means)/variance