loss function


Also found in: Wikipedia.

loss function

[′lȯs ‚fəŋk·shən]
(mathematics)
In decision theory, the function, dependent upon the decision and the true underlying distributions, which expresses the loss produced in taking the decision.
References in periodicals archive ?
Here, one is interested in learning (minimizing) a function over a labeled instance space, computed as an expectation of a loss function under an unknown distribution from which samples can be drawn.
The nominal rate is set by a central bank (ECB) that minimises a quadratic loss function (LF) for the monetary union, without taking into consideration financial stability:
Compared with [16], we adopt the [L.sub.2,q] norm instead of Frobenius norm in loss function to reduce the influence of noise and edge information.
Each branch learns task-specific features and has its own loss function corresponding to each task.
The main differences with the previous literature are: we only model the behavior of the CB and the inflation tax is the only source of financing, the objective of the CB is not to maximize social welfare but to minimize an ad hoc loss function similar to the one used in Barro and Gordon (1983), and the CB can only issue base money and nominal debt.
We use [L.sub.2,1] matrix norm based loss function as the reconstruction loss.
These methods require the use of a loss function, which measures the difference between the parameter and its estimator.
A general classifier is usually trained from supervised learning, which focuses on teaching the label information to the classifier and minimizes the designed loss function; therefore, it has learned specific feature mapping for each object.
Figures 13 and 14 compare and analyze the change of loss function value under different learning rates in the whole training process and make a brief analysis and explanation.
The propagation loss function represents the variation in that range, and the correlation distance will be short because the loss function has more fluctuation as distance change.
To train the DNN, we used the mean square error (MSE) loss function to measure the error between the predicted, and the truth vertex positions:
Assuming that the target value of Q is denoted by y, thus the loss function of Q-network is yielded: