Kelly Boothby

I had written a comment yesterday thinking I'd found a nice example differentiating the algorithms, but I was looking at the wrong plot and my comment was reaching a conclusion...

I do believe the ratios _should_ be bounded by a constant; I was too hasty in making that statement. The upper (single-pendant) plots show a ratio that appears to be...

Ah, good eye @dschult... sadly, the version of `bfs_f` in my repl doesn't have the double-removal problem so we don't get a performance boost for free :)

Thanks, this is due to an incompatible networkx version. I'm working on expanding our support window today.

After some thought, I wasn't a fan of my first approach; it was a bit too kludgy. Re-reading the comments of dwavesystems/minorminer#210, @pau557 had given me a better suggestion for...

Thanks for this report. In the code is an undocumented upper-bound calculated from weights and capacities, which is being violated in your example. ```python faux_inf = ( 3 * max(...

Sorry to duplicate your findings; I skipped to poking at code while sipping coffee this morning and forgot to read the bug report :roll_eyes: For what it's worth, I've minified...

As I'm staring at this, the pseudo-infinity is failing because it seems to assume that no pipe will carry more than any single consumer/producer. If we insert a `sum`, this...

and with that observation, we get a (provably?) minimal example ```python cap = 1000000000 def create_graph(with_capacity): G = nx.DiGraph() c = dict(capacity=cap) if with_capacity else {} # Add nodes with...

Pawing through the history a bit, the `faux_inf` bound was introduced in #1280 and has remained unchanged except for formatting since that time. I haven't yet read through the references...