Joseph Weston

WIP: add isosurface loss

1

This is the "simplest thing that could work" that I worked on with @akhmerov and @oroszl last week. Probably by thinking for 5 minutes we could come up with something...

WIP: replace BlockingRunner with AsyncRunner + asyncio.run

1

This eliminates a lot of code duplication. We also deprecate the use of BlockingRunner directly.

WIP: unify Learner1D, Learner2D and LearnerND

14

While writing the adaptive paper we managed to write down a simple algorithm formulated in terms of abstract domains and subdomains. Implementing such an algorithm in a learner should allow...

setup of test_ipyparallel_executor test is failing on windows on Python 3.8

2

See, for example, [here](https://dev.azure.com/python-adaptive/adaptive/_build/results?buildId=1306&view=logs&j=158e2689-29c5-5d1b-9cf0-c56b8d17aff2&t=ec5a3347-cde0-59b6-b623-38b4600252c4) It times out waiting for the string `"Engines appear to have started successfully"`. Perhaps `ipcluster` changed its output messages (though one would expect that they be...

See [here](https://dev.azure.com/python-adaptive/adaptive/_build/results?buildId=1306&view=logs&j=88437ce6-96ce-598b-fefb-346ab1c401a0&t=cd4a0a69-2aff-5683-8f13-fd1b0f42a2d9). The test was marked as "flaky". IMO this is a stop-gap solution and we should instead put a better error bound, for example [as shown here](http://www.ti3.tu-harburg.de/paper/rump/Ru11.pdf)

Add a deterministic Runner strategy

5

Currently if you use the default Runner with more than 1 concurrent task, your run will be nondeterministic because the concurrent tasks can finish in different orders. We could relatively...

improve isosurface computation

4

When computing the isosurface, when 4 simplex edges crosses the isosurface, we add 2 triangles to the isosurface triangulation. Right now we add these arbitrarily ([in these 3 lines](https://github.com/python-adaptive/adaptive/blob/5ca16a040c08f68c407aa301743e8bdba4cac759/adaptive/learner/learnerND.py#L1047-1049)) but...

Blocked by #238 Currently we use the "ad-hoc" package layout where the "adaptive" package is directly in the repository root. Such a layout can cause confusion when it comes to...

make runner.live_plot work with a wider range of backends

1

the docstring says: ``` plotter : function A function that takes the learner as a argument and returns a holoviews object. By default ``learner.plot()`` will be called. ``` But not...

If people are using the async runner then they have to jump through some hoops (`runner.task.print_stack()`) to find out what went wrong. It would be good if we could provide...