Todos los libros

Profesional

Apps Acerca de Coach Iniciar sesión Empezar a leer

Quantitative Finance · Glosario

¿Qué es Retraining schedule?

Definition 12.5 Machine Learning for Markets · Capítulo 12 — Online Learning and Drift

A retraining schedule is the rule that decides when a model is refitted and on which data: on a calendar (every week, on a rolling window), on a detector’s alarm, continuously (an online estimator), or by a combination; it is part of the model and is validated with it.

schedule (least squares on the last 500 observations unless stated)prequential R2R^2refits
never (fitted on the first 1 000 steps)−7.81%-7.81\%0
every 250 steps2.62%75
on a Page–Hinkley alarm (refit on the data since the alarm)−2.67%-2.67\%2
recursive least squares, λ=0.998\lambda = 0.9983.24%every step
the truth4.62%
Table 12.1. Four retraining schedules on a stream whose coefficients are redrawn every 2 000 steps on average (19 000 steps scored after a warm-up of 1 000). Data: ml_online.schedules.
Leer en el capítulo →