Transductive Learning with Multi-class Volume Approximation

Authors: Gang Niu, Bo Dai, Christoffel Plessis, Masashi Sugiyama

ICML 2014 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show it compares favorably with the one-vs-rest extension.In this section, we numerically evaluate MAVR.
Researcher Affiliation Collaboration Gang Niu NIUGANG@BAIDU.COM Tokyo Institute of Technology, Tokyo, 152-8552, Japan Baidu Inc., Beijing, 100085, China Bo Dai BODAI@GATECH.EDU Georgia Institute of Technology, Atlanta, GA 30332, USA Marthinus Christoffel du Plessis CHRISTO@SG.CS.TITECH.AC.JP Masashi Sugiyama SUGI@CS.TITECH.AC.JP Tokyo Institute of Technology, Tokyo, 152-8552, Japan
Pseudocode Yes Algorithm 1 MAVR Input: P, Q, Y , γ and τ Output: H and ρ 1: Eigen-decompose P and Q; 2: Construct the function g(ρ); 3: Find the smallest root of g(ρ); 4: Recover h using ρ and reshape h to H.
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 The latter 3 data sets come from Zelnik-Manor & Perona (2004).
Dataset Splits No The paper describes a transductive learning setting where prediction is made on unlabeled data (Xu), but it does not specify explicit dataset splits (e.g., percentages or counts) for training, validation, and testing needed for reproduction.
Hardware Specification No The paper does not provide specific hardware details (such as GPU or CPU models, or memory specifications) used for running the experiments.
Software Dependencies No The paper does not provide specific software dependencies with version numbers (e.g., library or framework versions) used for the experiments.
Experiment Setup Yes For the hyperparameters, we set γ = 99 and τ = l.The default values of factors were σϵ = 0.5, σ = 0.5, l = 3, n = 300, γ = 99, and τ = l, and the ranges of these factors were σϵ 0.5 exp{−1.5, −1.4, −1.3 . . . , 0.5}; l {3, 4, 5, . . . , 20}; n {120, 138, 156, . . . , 480}; σ 0.5 exp{−1, 0.9, 0.8, . . . , 1}; γ 99 exp{−4, −3, −2, . . . , 16}; l exp{−2, −1, 0, . . . , 18}.