Multi-Mode Deep Matrix and Tensor Factorization

Authors: Jicong Fan

ICLR 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The experiments on synthetic data and real datasets showed that the proposed methods have much higher recovery accuracy than many baselines.
Researcher Affiliation Academia Jicong Fan1,2 1School of Data Science, The Chinese University of Hong Kong (Shenzhen), China 2Shenzhen Research Institute of Big Data, China
Pseudocode Yes Algorithm 1 Gradient-based optimization for M2DMTF (12)
Open Source Code Yes Codes link: https://github.com/jicongfan/Multi-Mode-Deep-Matrix-and-Tensor-Factorization
Open Datasets Yes We consider two benchmark datasets: Movie Lens-100k and Movie Lens-1M... We compare the proposed method M2DMTF with the baselines on the following datasets: Amino acid fluorescence (Bro, 1997) (5 201 61), Flow injection (Nørgaard & Ridder, 1994) (12 100 89), and SW-NIR kinetic data (Bijlsma & Smilde, 2000) (301 241 8).
Dataset Splits No The paper mentions 'determine the hyper parameters of all methods via cross-validation' but does not explicitly state a separate validation dataset split with specific percentages or sample counts for the reported results in tables. The tables only show 'Train/Test' ratios.
Hardware Specification No No specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments were provided.
Software Dependencies No The paper mentions using MATLAB and Python, and refers to optimizers like 'i Rprop+' and 'Adam', but does not provide specific version numbers for any software or libraries.
Experiment Setup Yes In MF (problem (1) in the main paper), the factorization dimension d is 5 because it outperforms other choices. The λ is chosen from {0.01, 0.1, 1} and the optimizer is i Rprop+. The maximum iteration is 2000. ... In M2DMTF, L = 2, d1 = d2 = 3, h(1) 1 = h(2) 1 = 10, m1 = m2 = 20, and λ1 = λ 2 = 1. ... The activation function is the hyperbolic tangent function. The optimizer is i Rprop+ and the maximum iteration is 3000.