Hierarchical Clustering of Data Streams: Scalable Algorithms and Approximation Guarantees

Authors: Anand Rajagopalan, Fabio Vitale, Danny Vainstein, Gui Citovsky, Cecilia M Procopiuc, Claudio Gentile

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Finally, since our algorithms are principled but also very practical, we carry out an experimental comparison on both synthetic and real-world datasets showing competitive results against known baselines.
Researcher Affiliation Collaboration 1Google Research, NY, USA 2Lille University and INRIA Lille, France 3Tel-Aviv University, Israel.
Pseudocode Yes The general pseudocode for Insµ is given in Appendix A. In the following sections, we specify particular measures µ from which hyperplanes can be efficiently sampled and which additionally give rise to HC algorithms having the sequential property, and exhibiting good approximation ratios for the metrics of Section 2. The associated insertion operations are presented in the corresponding sections of the appendix.
Open Source Code No The paper does not provide any statements about open-sourcing its code or links to a code repository.
Open Datasets Yes For real-world datasets, we compare the algorithms on the following data of varying scale: MNIST, ALOI (Geusebroek et al. (2005)), and ILSVRC12 (Deng et al. (2009)) trained with Res Net34 architecture.
Dataset Splits No The paper describes generating synthetic data (10K examples from standard Gaussians) and sampling triplets for evaluation (10K triplets), but it does not specify explicit training, validation, or test dataset splits in terms of percentages, sample counts, or predefined splits for model development/evaluation in a traditional sense.
Hardware Specification No For hardware, we used machines with a maximum of 125GB of RAM and 16 CPUs. This provides general information but lacks specific models (e.g., CPU model, GPU model if any were used) or precise memory amounts, which are required for full reproducibility.
Software Dependencies No The paper does not specify any software dependencies with version numbers (e.g., programming languages, libraries, or specific solvers).
Experiment Setup Yes We choose σ as the mean ℓ2 distance between pairs of points. This is to ensure a reasonable distribution of similarity weights.