Joseph Weston

This seems complicated because we would need to convert arbitrary ReST to a markdown cell, but maybe there's some heuristic we can use

1. As long as the error is semantic (so we can use a fine-grained `except` to catch it) that's fine 2. On the command line you specify numeric IDs or...

tagging @ltalirz at his request.

Here's another datapoint: adding `gc.collect()` after `manager.unload_profile()` seems to solve the issue.

I've put this on my review backlog

Cool! We'll have a look, thanks.

_originally posted by Anton Akhmerov ([@anton-akhmerov](https://gitlab.kwant-project.org/anton-akhmerov)) at 2018-03-10T21:02:12.166Z on [GitLab](https://gitlab.kwant-project.org/qt/adaptive/issues/28)_ Depends on [gitlab:#50](https://gitlab.kwant-project.org/qt/adaptive/issues/50) (`priority_split` is the higher priority action based on the evaluation whether the error estimate is reliable).

I definitely agree that the loss should be larger for the case you showed. Would you agree that the reason why a loss of 1e-3 seems "wrong" is that the...

>someone starts from a bunch of data that does not necessarily include the boundaries IIRC in this case the learner will typically return the boundaries the next time you `ask`,...

e.g. This is a (pretty sick) example that uses [K3D](https://github.com/K3D-tools/K3D-jupyter)