Approximate Multiplication of Sparse Matrices with Limited Space

Authors: Yuanyu Wan, Lijun Zhang10058-10066

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we perform numerical experiments to verify the efficiency and effectiveness of our SCOD. Fig. 1 and 2 show the results of different algorithms among different ℓon the synthetic datasets.
Researcher Affiliation Academia Yuanyu Wan, Lijun Zhang National Key Laboratory for Novel Software Technology, Nanjing University, Nanjing 210023, China {wanyy, zhanglj}@lamda.nju.edu.cn
Pseudocode Yes Algorithm 1 Dense Shrinkage (DS), Algorithm 2 Simultaneous Iteration (SI), Algorithm 3 Verified Simultaneous Iteration (VSI), Algorithm 4 Sparse Co-occuring Directions (SCOD)
Open Source Code No The paper does not provide a specific repository link or an explicit statement about the release of source code for the described methodology.
Open Datasets Yes NIPS conference papers1 (Perrone et al. 2017) and Movie Lens 10M2. ... 1https://archive.ics.uci.edu/ml/datasets/NIPS+Conference+ Papers+1987-2015 2https://grouplens.org/datasets/movielens/10m/
Dataset Splits No The paper describes how the input matrices X and Y are derived from the original datasets (e.g., 'let XT be the first 2905 columns of M, and let Y T be the others'), but it does not specify train/validation/test splits for model evaluation.
Hardware Specification No The paper does not provide any specific details about the hardware used for running the experiments.
Software Dependencies No The paper mentions 'Matlab' as used for generating synthetic datasets but does not specify a version number or list other software dependencies with their versions for the implementation of the algorithm.
Experiment Setup Yes In all experiments, each algorithm will receive two matrices X Rmx n and Y Rmy n, and then output two matrices BX Rmx ℓand BY Rmy ℓ. We adopt the approximation error XY T BXBT Y and the projection error XY T π U(X)π V (Y )T to measure the accuracy of each algorithm, where U Rmx k, V Rmy k and we set k = 200. Furthermore, we report the runtime of each algorithm to verify the efficiency of our SCOD. Because of the randomness of SCOD, SFD-AMM, CS, RP and Hashing, we report the average results over 50 runs.