Algorithms for Caching and MTS with reduced number of predictions

Authors: Karim Ahmed Abdel Sadek, Marek Elias

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We perform an empirical evaluation of our caching algorithm F&R on the same datasets and with the same predictors as the previous works (Lykouris and Vassilvitskii, 2021; Antoniadis et al., 2023; Im et al., 2022).
Researcher Affiliation Academia Karim Abdel Sadek University of Amsterdam karim.abdel.sadek@student.uva.nl Marek Eliáš Department of Computing Sciences Bocconi University marek.elias@unibocconi.it
Pseudocode Yes Algorithm 1: Follower; Algorithm 2: Robustf (one phase)
Open Source Code Yes The code of our implementation can be found at https://github.com/marek-elias/caching/
Open Datasets Yes Bright Kite dataset (Cho et al., 2011) contains data from a certain social network. ... Citi Bike dataset contains data about bike trips in a bike sharing platform Citi Bike.
Dataset Splits No The paper describes dataset usage and selection criteria, such as "choose instances corresponding to the first 100 users with the longest check-in sequences" and "trimming length of each instance to 25 000", but does not specify explicit training, validation, or test dataset splits.
Hardware Specification No The paper does not provide specific details about the hardware (e.g., CPU, GPU models, memory) used for running the experiments.
Software Dependencies No The paper mentions implementing algorithms and using predictors, but does not provide specific version numbers for any software dependencies or libraries.
Experiment Setup Yes Notes on implementation of F&R. We follow the recommendations in Section 3 except that Follower switches to Robust whenever its cost is α = 1 times higher compared to Belady in the same period. With higher α, the performance of F&R approaches Ft P on the considered datasets. With k = 10 (Bright Kite dataset), we use F = [1, 6, 9] corresponding to f(i) = i. Note that, with such small k, polynomial and exponential f would also give a very similar F. With k = 100 (Citi Bike dataset), we use exponential f(i) = 2i+1 1. With a-separated queries, Follower uses LRU heuristic when prediction is unavailable, and Robust ignores F, querying the predictor at each page fault separated from the previous query by at least a time steps.