brainmass.objectives.timeseries_rmse

brainmass.objectives.timeseries_rmse()

Build a root-mean-square-error loss between two time series.

Returns:

loss(prediction, target) -> scalar computing sqrt(mean((prediction - target) ** 2)). The subtraction is unit-checked: incompatible units raise before the magnitude is taken.

Return type:

callable

See also

fc_rmse

RMSE between functional-connectivity matrices.

Examples

>>> import jax.numpy as jnp
>>> from brainmass import objectives
>>> loss = objectives.timeseries_rmse()
>>> x = jnp.zeros((10, 3))
>>> float(loss(x + 2.0, x))
2.0