# theanets.losses.MeanAbsoluteError¶

class theanets.losses.MeanAbsoluteError(target, weight=1.0, weighted=False, output_name='out')

Mean-absolute-error (MAE) loss function.

Notes

The mean absolute error (MAE) loss computes the mean difference between the output of a computation graph $$x = (x_1, \dots, x_d)$$ and its expected target value $$t = (t_1, \dots, t_d)$$. Mathematically,

$\begin{split}\begin{eqnarray*} \mathcal{L}(x, t) &=& \frac{1}{d} \|x - t\|_1 \\ &=& \frac{1}{d} \sum_{i=1}^d |x_i - t_i| \end{eqnarray*}\end{split}$

Whereas some MAE computations return the sum over dimensions, the MAE here is computed as an average over the dimensionality of the data.

For cases where $$x$$ and $$t$$ are matrices, the MAE computes the average over corresponding rows in $$x$$ and $$t$$.

__init__(target, weight=1.0, weighted=False, output_name='out')

Methods

 __init__(target[, weight, weighted, output_name]) log() Log some diagnostic info about this loss.

Attributes

 variables A list of Theano variables used in this loss.