penaltymodel-maxgap energy levels #76
Conversation
arcondello
force-pushed
the
fix/maxgap-energy-levels
branch
from
ede66a4
to
f78a14e
Compare
| @@ -57,6 +70,9 @@ def generate_ising(graph, feasible_configurations, decision_variables, | |||
| to a non-zero infeasible gap. | |||
|
|
|||
| """ | |||
| if len(graph) == 0: | |||
| return dimod.BinaryQuadraticModel.empty(dimod.SPIN), float('inf') | |||
There was a problem hiding this comment.
When MIP finds that it is given a table with no feasible configurations, it returns a zero-ed out BQM and a gap of zero. Here, MaxGap is returning an empty model and a gap of infinity. Perhaps we should have a consistent default gap when bad argument is given?
There was a problem hiding this comment.
Good idea. What do you think makes more sense?
There was a problem hiding this comment.
I like a gap of infinity because I think it's unlikely that a user would choose that as a gap. Hence, when you see it's an infinity you know it's an error or a computer generated default.
There was a problem hiding this comment.
I think this is out-of-scope for this PR then. See #79
* remove six dependency * refactor Theta class to use dimod's BQM object for logic * change generate_ising to generate which now returns a BQM * refactor unittests * fix offset bug
BQMs in which every configuration have a specified energy should be treated as having infinite gap. Reasoning: * It tests against gap>0 which is a good shorthand for a valid penalty model * The intuition is that if one increased energy from ground and stops when the first infeasible state is reached is maintained. In that case the energy would increase -> inf
arcondello
force-pushed
the
fix/maxgap-energy-levels
branch
from
88130d9
to
446b9fa
Compare
Codecov Report
@@ Coverage Diff @@
## master #76 +/- ##
=========================================
- Coverage 96.04% 95.94% -0.1%
=========================================
Files 26 28 +2
Lines 1389 1358 -31
=========================================
- Hits 1334 1303 -31
Misses 55 55
Continue to review full report at Codecov.
|
| Provides the structure for Theta. | ||
|
|
||
| linear_energy_ranges (dict): | ||
| A dict of the form {v: (min, max, ...} where min and max are the |
There was a problem hiding this comment.
Missing closing parethesis. v: (min, max)
Closes #73