Who Likes What? — SplitLBI in Exploring Preferential Diversity of Ratings

Authors: Qianqian Xu, Jiechao Xiong, Zhiyong Yang, Xiaochun Cao, Qingming Huang, Yuan Yao262-269

AAAI 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, three examples are exhibited with both simulated and real-world data to illustrate the validity of the analysis above and applications of the methodology proposed. The first example is with simulated data while the latter two exploit real-world data.
Researcher Affiliation Collaboration 1Key Lab. of Intelligent Information Processing, ICT, CAS, Beijing, China 2Tencent AI Lab, Shenzhen, Guangdong, China 3State Key Lab. of Information Security, IIE, CAS, Beijing, China 4School of Computer Science and Technology, UCAS, Beijing, China 5School of Cyber Security, UCAS, Beijing, China 6Key Lab. of Big Data Mining and Knowledge Management, CAS, Beijing, China 7Peng Cheng Laboratory, Shenzhen, Guangdong, China 8Department of Mathematics, HKUST, Hong Kong, China
Pseudocode Yes Algorithm 1 Sparse regularization path (...) Algorithm 2 Syn Par-Split LBI
Open Source Code Yes Because of the page limit, we only show the first two in the main body, while remaining the third one in the supplementary materials1. 1https://github.com/qianqianxu010/AAAI2020/tree/master/supplementary
Open Datasets Yes The Movie Lens 1M Data Set 2 is comprised of 3952 movies rated by 6040 users. (...) 2https://grouplens.org/datasets/movielens/
Dataset Splits Yes we randomly split the whole data samples into training set (70% of the total comparisons) and testing set (the remaining 30%). (...) To this end, we adopt a standard cross-validation scheme to choose the stopping time t: Given the training data, fix κ and α, then split the data S into S1, , SK, where Si Sj = φ, i = j, K i=1 Si = S.
Hardware Specification Yes Fig.1 (Left) shows the mean running time for 20 times repeat of Syn Par-Split LBI with thread number changing from 1 to 16 in a 16-core server with Intel(R) Xeon(R) E5-2670 2.60GHz CPU and 384GB of RAM. The server runs Linux 4.2.0 64bit.
Software Dependencies No The paper mentions 'Linux 4.2.0 64bit' but does not specify other ancillary software dependencies (e.g., programming languages, libraries, frameworks) with specific version numbers.
Experiment Setup Yes Simulated Study Settings. We validate the proposed algorithm on simulated data with n = |V | = 50 labeled by 100 users. Specifically, we first generate the feature matrix for each node: X = [X i ]n i=1 Rn d, where Xi is a d-dimensional (d = 20 in this experiment) column feature vector drawn randomly from N(0, 1) representing node i. Then each entry of the common coefficient β has a probability p1 = 0.4 with nonzero value and they are drawn randomly from N(0, 1). Besides, for each user u, each entry of his personalized deviation coefficient δu has a probability p2 = 0.4 to be nonzero and is drawn randomly from N(0, 1). (...) Input: Data (X, y), damping factor κ, step size α.