BalancingLearner puts all points in the first child-learner when asking for points with no data present

[basnijholt]

def h(x, offset=0):
 a = 0.01
 return x + a**2 / (a**2 + (x - offset)**2)

learners = [adaptive.Learner1D(partial(h, offset=random.uniform(-1, 1)),
 bounds=(-1, 1)) for i in range(10)]

bal_learner = adaptive.BalancingLearner(learners)

bal_learner.ask(10)

([(0, -1),
 (0, 1),
 (0, 0.0),
 (0, -0.5),
 (0, 0.5),
 (0, -0.75),
 (0, -0.25),
 (0, 0.25),
 (0, 0.75),
 (0, -0.875)],
 [inf, inf, inf, inf, inf, inf, inf, inf, inf, inf])

Instead, it should balance the points over all learners. This happens because the loss improvements are inf for all points.

A temporary solution would be to force all learners to have the same amount of points by using:

bal_learner = adaptive.BalancingLearner(learners, strategy='npoints')