Spectral Bandits for Smooth Graph Functions

Authors: Michal Valko, Remi Munos, Branislav Kveton, Tomáš Kocák

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments on real-world content recommendation problem show that a good estimator of user preferences for thousands of items can be learned from just tens of nodes evaluations. (Abstract) and 6. Experiments (Section Title)
Researcher Affiliation Collaboration Michal Valko MICHAL.VALKO@INRIA.FR INRIA Lille Nord Europe, Seque L team, 40 avenue Halley 59650, Villeneuve d Ascq, France Remi Munos REMI.MUNOS@INRIA.FR INRIA Lille Nord Europe, Seque L team, France; Microsoft Research New England, Cambridge, MA, USA Branislav Kveton BRANISLAV.KVETON@TECHNICOLOR.COM Technicolor Research Center, 735 Emerson St, Palo Alto, CA 94301, USA Tom aˇs Koc ak TOMAS.KOCAK@INRIA.FR INRIA Lille Nord Europe, Seque L team, 40 avenue Halley 59650, Villeneuve d Ascq, France
Pseudocode Yes Algorithm 1 SPECTRALUCB and Algorithm 2 SPECTRALELIMINATOR
Open Source Code No The paper does not provide an explicit statement or link for the open-source code of the described methodology.
Open Datasets Yes Movie Lens dataset (Lam & Herlocker, 2012), a dataset of 6k users who rated one million movies. (Section 6.2) and The social network of the users was crawled by Jamali & Ester (2010) (Section 6.3)
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, or explicit mention of training, validation, and test sets) needed to reproduce the data partitioning in a standard ML context.
Hardware Specification No The paper does not provide specific hardware details (e.g., exact GPU/CPU models, processor types, or memory amounts) used for running its experiments.
Software Dependencies No The paper mentions using 'fast SDD solvers as CMG by Koutis et al. (2011)' but does not provide a specific version number for CMG or any other software dependency.
Experiment Setup Yes In all experiments we set δ to 0.001 and R to 0.01. For SPECTRALUCB and SPECTRALELIMINATOR we set Λ to ΛL + λI with λ = 0.01. For Lin UCB we regularized with λI with λ = 1. (Section 6. Experiments) and We ran the algorithms in the desired T < N regime, with N = 500 (N = 54 for the lattice), T = 250, and k = 5. (Section 6.1 Random graph models)