Adversarially Robust Dense-Sparse Tradeoffs via Heavy-Hitters

Authors: David Woodruff, Samson Zhou

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

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we describe our empirical evaluations for comparing the flip number of the entire vector and the flip number of the residual vector on real-world datasets.
Researcher Affiliation Academia David P. Woodruff Department of Computer Science Carnegie Mellon University dwoodruf@andrew.cmu.edu Samson Zhou Department of Computer Science Texas A&M University samsonzhou@gmail.com
Pseudocode Yes Algorithm 1 ROBUSTHH: Adversarially robust Lp-heavy hitters
Open Source Code Yes Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [Yes] Justification: Yes, open access to the data and code are provided in the full version of the paper.
Open Datasets Yes CAIDA traffic monitoring dataset. We used the CAIDA dataset [CAI16] of anonymized passive traffic traces from the equinix-nyc data center s high-speed monitor.
Dataset Splits No The paper describes an empirical evaluation comparing flip numbers but does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) for training, validation, or testing.
Hardware Specification Yes Our empirical evaluations were performed Python 3.10 on a 64-bit operating system on an AMD Ryzen 7 5700U CPU, with 8GB RAM and 8 cores with base clock 1.80 GHz.
Software Dependencies No The paper mentions 'Python 3.10' but does not list multiple key software components with their versions or a self-contained solver/package with a specific version number as required for a 'Yes' answer.
Experiment Setup Yes We compare the flip number of the entire data stream versus the flip number of the residual vector across various values of the algorithm error ε {10 1, 10 2, . . . , 10 5}, values of the heavy-hitter threshold α {4 1, 4 2, . . . , 4 10}, and the frequency moment parameter p {1.1, 1.2, . . . , 1.9}.