Temporal Regularized Matrix Factorization for High-dimensional Time Series Prediction

Authors: Hsiang-Fu Yu, Nikhil Rao, Inderjit S. Dhillon

NeurIPS 2016 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Experimental results show the superiority of TRMF in terms of scalability and prediction quality. In particular, TRMF is two orders of magnitude faster than other methods on a problem of dimension 50,000, and generates better forecasts on real-world datasets such as Wal-mart E-commerce datasets. We demonstrate the superiority of the proposed approach via extensive experimental results in Section 5
Researcher Affiliation Collaboration Hsiang-Fu Yu University of Texas at Austin rofuyu@cs.utexas.edu Nikhil Rao Technicolor Research nikhilrao86@gmail.com Inderjit S. Dhillon University of Texas at Austin inderjit@cs.utexas.edu
Pseudocode No The paper describes the mathematical formulations and optimization steps for TRMF, including updates for F, X, and W, but it does not provide any explicitly labeled 'Pseudocode' or 'Algorithm' blocks.
Open Source Code No The paper does not provide any statement about releasing source code for the methodology, nor does it include links to a code repository.
Open Datasets Yes The data sets electricity and traffic are obtained from the UCI repository, while walmart-1 and walmart-2 are two propriety datasets from Walmart E-commerce containing weekly sale information.
Dataset Splits Yes For each method and data set, we perform a grid search over various parameters (such as k, λ values) following a rolling validation approach described in [11].
Hardware Specification No The paper discusses computational scalability and efficiency (e.g., 'two orders of magnitude faster', 'O(|L|Tk2) time'), but it does not specify any particular hardware used for running the experiments (e.g., CPU type, GPU model, memory).
Software Dependencies No The paper mentions software like the 'R-DLM package [12]' but does not provide specific version numbers for any software dependencies or libraries used in their implementation (e.g., 'Python 3.x', 'PyTorch 1.x').
Experiment Setup Yes For L, we use {1, 2, . . . , 8} for synthetic, {1, . . . , 24}[{7 24, . . . , 8 24 1} for electricity and traffic, and {1, . . . , 10}[ {50, . . . , 56} for walmart-1 and walmart-2 to capture seasonality. For each method and data set, we perform a grid search over various parameters (such as k, λ values) following a rolling validation approach described in [11].