BIZNIQUE / LINEAR ALGEBRA / 11
Put feature scales in context
Put feature scales in context
See how units can dominate a distance even when neither feature is more meaningful.
Embedded Files
THE IDEA
Read the example.
Read the example.
Standardisation subtracts a training-set mean and divides by a training-set standard deviation. It changes the geometry seen by distance-sensitive models and can improve the conditioning of an optimisation problem.
zⱼ = (xⱼ − μⱼ) / sⱼ
Use it well
Use it well
Fit scaling parameters on training data only. Handle constant columns explicitly; dividing by a zero standard deviation is undefined.
TRY IT YOURSELF
A quick check.
A quick check.
If a training feature has mean 10 and standard deviation 2, what is the standardised value of 14?
Check your reasoning
Check your reasoning
(14 − 10)/2 = 2.
Page updated
Google Sites
Report abuse