Extending Gossip Algorithms to Distributed Estimation of U-statistics
Authors: Igor Colin, Aurélien Bellet, Joseph Salmon, Stéphan Clémençon
NeurIPS 2015 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | numerical experiments provide empirical evidence the proposed algorithms surpasses the previously introduced approach. Experiments conducted on AUC and within-cluster point scatter estimation using real data confirm the superiority of our approach. Section 5 presents our numerical results. |
| Researcher Affiliation | Academia | LTCI, CNRS, T el ecom Paris Tech Universit e Paris-Saclay 75013 Paris, France first.last@telecom-paristech.fr Aur elien Bellet Magnet Team INRIA Lille Nord Europe 59650 Villeneuve d Ascq, France aurelien.bellet@inria.fr |
| Pseudocode | Yes | Algorithm 1 Go Sta-sync: a synchronous gossip algorithm for computing a U-statistic. Algorithm 2 Go Sta-async: an asynchronous gossip algorithm for computing a U-statistic. |
| Open Source Code | No | The paper does not provide any statements or links regarding the availability of source code for the methodology. |
| Open Datasets | Yes | This dataset is available at http://mldata.org/repository/data/viewslug/svmguide3/ This dataset is available at https://archive.ics.uci.edu/ml/datasets/Wine |
| Dataset Splits | No | The paper describes the datasets and their properties but does not specify training, validation, or test splits in the traditional machine learning sense for model training, as the algorithms aim to estimate a U-statistic over the entire distributed dataset. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments. |
| Software Dependencies | No | The paper does not specify any software dependencies or their version numbers. |
| Experiment Setup | Yes | We perform our simulations on the three types of network described below (corresponding values of 1 λ2(2) are shown in Table 1). Here, we use k = 5 and p = 0.3 to achieve a connectivity compromise between the complete graph and the two-dimensional grid. For each generated network, we perform 50 runs of Go Sta-sync (Algorithm 1) and U2-gossip. |