All translations
Enter a message name below to show all available translations.
Found 3 translations.
| Name | Current message text |
|---|---|
| h English (en) | * '''Constant''' — simple but may overshoot or stall. * '''{{Term|step decay}}''' — multiply <math>\eta</math> by a factor (e.g. 0.1) every <math>k</math> {{Term|epoch|epochs}}. * '''{{Term|exponential decay}}''' — <math>\eta_t = \eta_0 \, e^{-\lambda t}</math>. * '''{{Term|cosine annealing}}''' — smoothly reduces the rate following a cosine curve, often with warm restarts. * '''Linear {{Term|learning rate warmup|warm-up}}''' — ramp up from a small <math>\eta</math> during the first few {{Term|iteration|iterations}} to stabilise early training. |
| h Spanish (es) | * '''Constante''' — simple, pero puede sobrepasar o estancarse. * '''{{Term|step decay|Decaimiento por pasos}}''' — multiplicar <math>\eta</math> por un factor (p. ej., 0.1) cada <math>k</math> {{Term|epoch|épocas}}. * '''{{Term|exponential decay|Decaimiento exponencial}}''' — <math>\eta_t = \eta_0 \, e^{-\lambda t}</math>. * '''{{Term|cosine annealing|Recocido coseno}}''' — reduce la tasa de forma suave siguiendo una curva coseno, a menudo con reinicios cálidos. * '''{{Term|learning rate warmup|Calentamiento}} lineal''' — incrementa <math>\eta</math> desde un valor pequeño durante las primeras {{Term|iteration|iteraciones}} para estabilizar el entrenamiento inicial. |
| h Chinese (zh) | * '''常数''' — 简单但可能越过最优值或停滞。 * '''{{Term|step decay|阶梯衰减}}''' — 每 <math>k</math> 个 {{Term|epoch|轮次}} 将 <math>\eta</math> 乘以一个因子(例如 0.1)。 * '''{{Term|exponential decay|指数衰减}}''' — <math>\eta_t = \eta_0 \, e^{-\lambda t}</math>。 * '''{{Term|cosine annealing|余弦退火}}''' — 沿余弦曲线平滑地降低学习率,通常配合热重启使用。 * '''线性{{Term|learning rate warmup|预热}}''' — 在最初的几次{{Term|iteration|迭代}}中将 <math>\eta</math> 从一个较小的值逐步提升,以稳定早期训练。 |