Reconstructing Hidden Permutations Using the Average-Precision (AP) Correlation Statistic

Authors: Lorenzo De Stefani, Alessandro Epasto, Eli Upfal, Fabio Vandin

AAAI 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We complement our theoretical analysis with extensive experiments showing that unsupervised methods based on our model can precisely identify ground-truth clusters of rankings in real-world data. In particular, when compared to the Kendall s tau based methods, our methods are less affected by noise in low-rank items. We experimentally evaluate our model and algorithms with both synthetic and real-world data.
Researcher Affiliation Academia Lorenzo De Stefani, Alessandro Epasto, Eli Upfal Brown University Providence, RI 02906, United States {lorenzo, epasto, eli}@cs.brown.edu Fabio Vandin University of Padova Padova, PD 35131, Italy vadinfa@dei.unipd.it
Pseudocode No The paper describes the 'MAP (β, π) Generative Process' in a step-by-step manner, but it is not formally labeled as 'Pseudocode' or 'Algorithm' and lacks typical pseudocode formatting.
Open Source Code No The paper does not provide an explicit statement or link for the open-source code of the methodology described.
Open Datasets Yes We split each dataset in training and testing sets using 5-fold cross-validation. We applied the following simple classification algorithm. For each class we reconstructed the center permutation using our comparison-based algorithm (in Section 5.2) on the set of permutations belonging to that class in the training set.
Dataset Splits Yes We split each dataset in training and testing sets using 5-fold cross-validation.
Hardware Specification No The paper states 'Each run used a single core and less than 4GB of RAM' but does not specify any hardware models (e.g., CPU, GPU) or processing units.
Software Dependencies No The paper states 'We implemented the algorithms in C++' but does not provide specific version numbers for C++ compilers or any other software libraries used.
Experiment Setup Yes In this experiment, we generated, using the algorithm defined in Section 4, a set P of i.i.d. permutations of size n = 100 from the MAP (β, π) model with different settings of β and size of |P|. We set α = 0.15 as the jump-back probability in all the walks. We selected t = 200 u.a.r. nodes for each of the eight well-represented categories in the Wikipedia taxonomy. We split each dataset in training and testing sets using 5-fold cross-validation.