diff --git a/paper.md b/paper.md index 31c4af94e4f9c62b93b8715d6ef5d809cc444b2a..8dd5c35a515978e70c5e7be55d21ca4abd0ebae8 100755 --- a/paper.md +++ b/paper.md @@ -323,19 +323,18 @@ runner = Runner(learner, goal) Again, like the `Learner1D`, it is possible to specify a custom loss function. For example, the loss function used to find the iso-line in Fig. @fig:isoline (b) is ```python -def isoline_loss_function(level, sigma, priority): - from adaptive.learner.learnerND import default_loss +from adaptive.learner.learnerND import default_loss - def gaussian(x, mu, sigma): - return np.exp(-(x - mu) ** 2 / sigma ** 2 / 2) +def gaussian(x, mu, sigma): + return np.exp(-(x - mu) ** 2 / sigma ** 2 / 2) +def isoline_loss_function(level, sigma, priority): def loss(simplex, values, value_scale): values = np.array(values) dist = abs(level * value_scale - values).mean() L_default = default_loss(simplex, values, value_scale) L_dist = priority * gaussian(dist, 0, sigma) return L_dist + L_default - return loss loss_per_simplex = isoline_loss_function(0.1, 0.4, 0.5)