ratioscaled

Ratioscaled is an adjective used to describe data, measurements, or representations that have been transformed through a scaling operation based on a ratio between a target scale and an original scale. The term emphasizes that the transformation preserves proportional relationships by multiplying quantities by a constant or by a set of ratios across components.

The concept sits at the intersection of normalization and unit conversion. In practice, ratio scaling is employed to harmonize disparate data sources, convert units, or prepare inputs for algorithms that assume consistent ranges. It is also used in graphics and computer graphics to resize coordinates while preserving aspect relationships.

There are two common variants. Uniform ratio scaling applies a single scale factor across all components, x' = x × k, making it possible to bring measurements into a desired range such as 0 to 1. Feature-wise ratio scaling uses a distinct factor for each component, x_i' = x_i × k_i, enabling alignment of distributions or units that vary by feature.

Advantages of ratio-scaled transformations include preserving relative proportions and straightforward implementation. Limitations include potential distortion if the chosen ratio does not reflect the intended target, and the need to recalculate when the target scale changes.

See also: normalization, unit conversion, affine transformation, scaling.