Multilinear Regression for Embedded Feature Selection with Application to fMRI Analysis

Authors: Xiaonan Song, Haiping Lu

AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experiments on synthetic and real data show the advantages of Remurs compared to Lasso, Elastic Net, and their multilinear extensions.
Researcher Affiliation Academia Xiaonan Song Department of Computer Science Hong Kong Baptist University, China sxnzxz@gmail.com Haiping Lu Department of Computer Science University of Sheffield, UK h.lu@sheffield.ac.uk
Pseudocode Yes Algorithm 1 The Remurs algorithm based on ADMM.
Open Source Code No The paper does not contain any explicit statement about releasing the source code or a link to a code repository for the described methodology.
Open Datasets Yes We perform real-world f MRI classification on the CMU2008 dataset (Mitchell et al. 2008), with 3D f MRI data of size 51 61 23 (71,553 voxels). It aims to predict human brain activity associated with the meanings of nouns. The data acquisition experiments had nine right-handed subjects who viewed 60 different word-picture stimuli from 12 semantic categories, with 5 exemplars per category and 6 runs per stimulus. The numbers of valid brain voxels range from 19,750 to 21,764. Data were preprocessed with the SPM software and we use the preprocessed 3D data available online,1 where each voxel feature is the respective mean percent signal change (PSC) value over time. 1http://www.cs.cmu.edu/afs/cs/project/theo73/www/science2008/data.html
Dataset Splits Yes We follow (Kampa et al. 2014) to arrange the test, validation and training sets in the format of (1:1:4) for the six runs in all the experiments of Table 1, and report the average results.
Hardware Specification No The paper does not provide specific details regarding the hardware (e.g., CPU, GPU models, memory, or cloud instances) used for running the experiments.
Software Dependencies No The paper mentions 'SPM software' for data preprocessing but does not provide specific version numbers for any software dependencies or libraries.
Experiment Setup Yes Hyperparameters of all the methods are determined via fourfold cross validation on the training set, with range {10 3, 5 10 3, 10 2, 5 10 2, . . . , 5 102, 103}. In addition, γ and τ of Remurs are constrained by reasonable feature numbers (5% to 50% of brain voxels) in cross validation. ... we fix ρ to 1 in implementation.