| paper | d4-conf-ijcai-2018-0353 |
| authors | — |
| venue | ? |
| filed under | unclassified |
| judged by | gpt-6-luna / xhigh (triple__luna__xhigh__c6r1) |
| judge confidence | high |
| authors would recognise it | unclear |
The paper describes four heuristics and reports empirical comparisons, but contains no numbered theorem, lemma, corollary, or proposition asserting a computational result eligible to anchor a mirror. The proposed minimax schedule problem is a plausible follow-on, but it does not provide a continuous analogue of a qualifying result in this paper.
fails bit a — no named computational result to mirror
No named computational result is covered; the proposed problem does not mirror the four heuristics' empirical comparisons or establish a guarantee for them.
The paper reports four heuristic algorithms and empirical fairness benchmarks, but it has no numbered theorem, lemma, corollary, or proposition asserting a computational result such as tractability or hardness. Algorithms 1–4 and Tables 1–3 do not provide an eligible anchor under your criteria. So I can’t honestly make an anchor-based case for a continuous mirror.
There is still a plausible continuous problem in the paper’s allocation domain. Call it Continuous Minimax-Comfort Load Shedding. Consider a utility serving many households drawn from a modest set \(T\) of complete profile types. A type \(t\) includes a household’s hourly demand \(\ell_{t,h}\) and comfort cost \(\delta_{t,h}\); its mass \(\mu_t\) is the fraction of customers with that profile. For a horizon of \(H\) hours, a household’s outage schedule is \(z\in\{0,1\}^H\), where \(z_h=1\) means disconnected. Let \(y_{t,z}\) be the fraction of the population assigned schedule \(z\). Given per-capita supply \(g_h\), require
\[
\sum_z y_{t,z}=\mu_t
\quad\text{for each }t,
\qquad
\sum_{t,z} y_{t,z}(1-z_h)\ell_{t,h}\le g_h
\quad\text{for each }h.
\]
The discomfort of a type-\(t\) household on schedule \(z\) is \(d_t(z)=\sum_h z_h\delta_{t,h}\). The problem is to minimize \(q\), the greatest discomfort borne by any household, with \(y_{t,z}=0\) whenever \(d_t(z)>q\). A solution assigns mass to schedules; each household still receives an integral outage schedule.
A credible regime would be a large utility serving many households that share recurring appliance and usage profiles. The number of customers could be very large while the number of complete profiles \(\tau=|T|\) stays moderate. That is plausible for load planning, though the paper’s own comfort profiles are individualized enough that this compression is not automatic.
For variable \(H\), I would expect the question’s difficulty to turn on the schedule-pricing problem; the discomfort bound makes that problem knapsack-like, so continuum-specific hardness is a possibility, not a result. For fixed \(H\), the schedule set is finite and an explicit LP is polynomial in \(\tau\), though it may be impractically large.
The weak point is decisive: this is a well-motivated continuous version of the paper’s fairness objective, but it is a new optimization problem, not a mirror of a named result in the paper. The authors might recognize the household-level allocation model and comfort criterion; they would not be entitled to claim that the paper’s reported heuristic comparisons establish anything about its complexity.
The strongest objection is that this paper gives no computational result to mirror. Its numbered algorithms are heuristics, and its tables report empirical comparisons; it proves no complexity bound for a defined optimization problem. The proponent’s proposed Continuous Minimax-Comfort Load Shedding is therefore a new optimization problem, not a continuous version of a result in this paper. That is a decisive gap for calling this a worthwhile *mirror*.
The proposed problem is coherent: a large utility could have many households sharing a modest number of demand and comfort profiles. The paper’s 372 observed households do not establish that regime—their hourly profiles are detailed and household-specific—but a larger utility with recurring profiles is plausible. So the negative case should not rest on claiming that household populations cannot have high multiplicity.
The harder question is whether the proposed model preserves the paper’s problem. It replaces a comparison of four policies across several fairness criteria with a planner that assigns whole-horizon outage schedules to minimize worst-case discomfort. That is a natural follow-on, and the authors might recognize it. But it is not an analysis of the reported heuristics or a mirror of a computational claim the paper makes. To capture their recurring parity goal, the model would also need to track each household’s accumulated disconnections or discomfort. Those histories can split households that began with the same profile into different types, potentially eroding the small-support premise over repeated planning periods. Dropping history keeps the type space compact, but changes the fairness question.
The proposed LP also introduces a variable horizon \(H\) and its schedule choices. With \(H\) fixed to an operational day or week, the schedule set has \(2^H\) possibilities; treating \(H\) as variable makes temporal scheduling the central combinatorial dimension. Either way, this may be useful work on fair load scheduling, but its main computational question is introduced by the new formulation rather than inherited from a named result in the paper.
So the case against is strongest as a case against this paper as a ChoCo mirror: there is no computational theorem to carry across, and the plausible continuous model changes the task. The broader claim that no scenario could yield worthwhile continuous work is weaker. The schedule-assignment problem is sensible and could be worth studying on its own; the supplied text does not justify treating it as a mirror of the paper’s results.
The adversarial triple: the proponent anchors on up to three named results; the opponent sees that case and must defeat every anchor; the judge decides which case convinced it. These are the pipeline’s own outputs, generated by tools/triple_run.py — no human edited them. The paper’s own text is not reproduced here beyond the quoted statement above.