Scheduling with Untrusted Predictions

Authors: Evripidis Bampis, Konstantinos Dogeas, Alexander Kononov, Giorgio Lucarelli, Fanny Pascual

IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In addition, we confirm the above theoretical bounds by conducting experimental evaluation comparing the proposed algorithms to the offline optimal ones for both the single and multiple machines settings.
Researcher Affiliation Academia 1Sorbonne Universit e, CNRS, LIP6, F-75005 Paris, France 2Novosibirsk State University, Novosibirsk, Russia 3Sobolev Institute of Mathematics, Novosibirsk, Russia 4Universit e de Lorraine, LCOMS, Metz, France
Pseudocode No The paper describes algorithms such as SRPPT and SPPT(m) in prose, accompanied by lemmas and theorems, but does not present them in a structured pseudocode block or a clearly labeled 'Algorithm' section with step-by-step instructions.
Open Source Code Yes The code is publicly available at: https://github.com/ildoge/SUP.
Open Datasets No We create artificial instances, each one consisting of n = 50 jobs for the single machine case, following the same approach as in [Purohit et al., 2018]. We draw the actual processing time values, xj, independently for each job from a Pareto distribution with a parameter α = 1.1.
Dataset Splits No The paper describes generating artificial instances and performing independent runs, but it does not specify explicit training, validation, and test dataset splits with percentages or sample counts.
Hardware Specification No The paper does not provide specific details about the hardware used for running the experiments, such as CPU/GPU models, memory, or cloud instance types.
Software Dependencies No The paper mentions implementing algorithms but does not specify any software dependencies or their version numbers (e.g., specific programming languages, libraries, or solvers with versions).
Experiment Setup Yes We use values of σ that belong in [0, 5000] using a step of 50. The performance of the algorithms is compared to the optimal Shortest Remaining Processing Time First (SRPT) algorithm using the actual processing values. We set τ = 0.1 for the first simulations. ... with λ = 1/2, which performs well for small values of error while remaining competitive to Round Robin when the error is big.