Where does loss collide?

In electrical engineering, loss is dissipated energy: the part of the input energy that turns into heat and does no useful work, such as the energy spent each cycle remagnetising an iron core. Steinmetz's 1892 On the law of hysteresis first gave an empirical law for this hysteresis loss, relating the energy lost per cycle to the peak magnetic flux density. The quantity is central to the efficiency of motors, transformers, and inductors. The referent is physical energy turned to heat inside a material, not surplus destroyed in a market, a cost assigned to a wrong estimate, or an asymmetry in how people value outcomes.[1]

Unit

Energy converted to heat per cycle or per unit time, in joules or watts

Recovery means

Not applicable

Measured by

Found by calorimetry or from the area of the hysteresis loop, and reported as a loss coefficient

In statistical decision theory, loss is the loss function: a rule that assigns a numerical cost to each possible wrong decision or estimation error, so a procedure can be judged by the expected cost it incurs. Wald's 1939 Contributions to the theory of statistical estimation and testing hypotheses introduced this function, later developed fully in his 1950 Statistical decision functions; the familiar squared-error criterion descends from the older method of least squares. The referent is a mathematical penalty attached to error, chosen by the analyst, not heat lost in a material, welfare destroyed by a distortion, or a felt aversion to losing.[2][3]

Unit

A numerical cost assigned to a decision or an estimation error

Recovery means

Not applicable

Measured by

Specified by the analyst as a loss function, then summarised by its expected value, the risk