Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].

A (1+ε)-Approximation for Ultrametric Embedding in Subquadratic Time

Authors: Gabriel Bathie, Guillaume Lagarde

AAAI 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental 1.2 Our Contribution Experimental Results To complement our theoretical results and to demonstrate the practical efficiency of our algorithm, we perform an extensive set of experiments. We measure the performance of our algorithm both in terms of approximation factor and running time on five classical and diverse real-world datasets, and evaluate its scalability on large synthetic datasets.
Researcher Affiliation Academia Gabriel Bathie1, 2, Guillaume Lagarde1 1 La BRI, Universit e de Bordeaux, France 2 DI ENS, PSL Research University, Paris, France
Pseudocode Yes Algorithm 1: γ α-approx. of the best ultrametric fit...Algorithm 2: α-approximation of the cut weights
Open Source Code No Our algorithm is implemented in the Rust programming language, version 1.79.0 (129f3b996, 2024-06-10). Our code was compiled in release mode. The paper mentions the implementation language and version, but does not provide concrete access to the source code (e.g., a repository link or an explicit statement of public availability).
Open Datasets Yes All datasets are publicly available on the UCI ML Repository (Kelly, Longjohn, and Nottingham 2024b), or Kaggle (Kelly, Longjohn, and Nottingham 2024a) for the DIABETES dataset.
Dataset Splits No We measure the performance of our algorithm both in terms of approximation factor and running time on five classical and diverse real-world datasets, and evaluate its scalability on large synthetic datasets. The paper mentions using real-world and synthetic datasets for evaluation but does not specify any training, validation, or test splits. The experimental setup describes running the algorithm multiple times on the datasets, not splitting them for model training/evaluation.
Hardware Specification Yes The hardware configuration includes an Intel(R) Xeon(R) CPU E5-2630 v3 @ 2.40GHz and 126GB of RAM.
Software Dependencies Yes our algorithm is implemented in the Rust programming language, version 1.79.0 (129f3b996, 2024-06-10).
Experiment Setup Yes We measure the performance of our BUF c-approximation algorithm using our γ-KT and α-ACW algorithms with α = γ = c, which we call FASTULT. ... We evaluate FASTULT for different values of c, and for each value we run the algorithm t = 30 times on each of the 5 datasets, for a total of 150 runs per value of c.