Skip to content

feat(gui): three research tabs (energy landscape, quantum convergence… - #9

Merged
jaewonyun1234 merged 1 commit into
mainfrom
feat/synthetic-only-animated-map
Jun 23, 2026
Merged

jaewonyun1234 merged 1 commit into
mainfrom
feat/synthetic-only-animated-map

Conversation

@jaewonyun1234

Copy link
Copy Markdown
Owner

…, hardness)

Replace the dead "live CP-SAT convergence" tab (CP-SAT one-shots these instances in ~20ms / 1 incumbent, so there was nothing to animate) with three research-grade panels, backed by a tested pure-function analysis layer (contrail_env/analysis.py):

  • Energy landscape: histogram of every feasible solution's cost with the optimum marked, overlaid with the costs GBS and uniform-random samplers actually achieve -- does the quantum sampler concentrate near the optimum?
  • Quantum convergence: best-cost-so-far vs work for Pasqal QAOA and Xanadu GBS on the current instance, against the CP-SAT optimum -- a curve that actually moves (these iterate).
  • Hardness sweep: CP-SAT wall time + incumbents vs problem size -- shows the instances are trivially easy for the exact solver (incumbents stays flat).

analysis.py is unit-tested (landscape optimum matches CP-SAT; repaired samples never beat the optimum; sweep shapes/effort). GUI plotting slots smoke-tested headless. Tabs renumber to 8; the SolveWorker progress curve is gone.

…, hardness)

Replace the dead "live CP-SAT convergence" tab (CP-SAT one-shots these
instances in ~20ms / 1 incumbent, so there was nothing to animate) with three
research-grade panels, backed by a tested pure-function analysis layer
(contrail_env/analysis.py):

- Energy landscape: histogram of every feasible solution's cost with the
  optimum marked, overlaid with the costs GBS and uniform-random samplers
  actually achieve -- does the quantum sampler concentrate near the optimum?
- Quantum convergence: best-cost-so-far vs work for Pasqal QAOA and Xanadu GBS
  on the current instance, against the CP-SAT optimum -- a curve that actually
  moves (these iterate).
- Hardness sweep: CP-SAT wall time + incumbents vs problem size -- shows the
  instances are trivially easy for the exact solver (incumbents stays flat).

analysis.py is unit-tested (landscape optimum matches CP-SAT; repaired samples
never beat the optimum; sweep shapes/effort). GUI plotting slots smoke-tested
headless. Tabs renumber to 8; the SolveWorker progress curve is gone.
@jaewonyun1234
jaewonyun1234 merged commit 710b7fc into main Jun 23, 2026
1 check passed
@jaewonyun1234
jaewonyun1234 deleted the feat/synthetic-only-animated-map branch June 23, 2026 22:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant