fastFM: A Library for Factorization Machines

Authors: Immanuel Bayer

JMLR 2016 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments show, that the ALS and MCMC solver in fast FM compare favorable to lib FM with respect to runtime (Figure 2) and are indistinguishable in terms of accuracy. The experiments have been conducted on the Movie Lens 10M data set using the original split with a fixed number of 200 iterations for all experiments.
Researcher Affiliation Academia Immanuel Bayer EMAIL University of Konstanz 78457 Konstanz , Germany
Pseudocode No The paper includes Python code snippets demonstrating usage, but no formal pseudocode or algorithm blocks are present.
Open Source Code Yes fast FM contains a test suite that is run on each commit to the Git Hub repository via a continuous integration server4. https://travis-ci.org/ibayer/fast FM-core
Open Datasets Yes The experiments have been conducted on the Movie Lens 10M data set using the original split with a fixed number of 200 iterations for all experiments.
Dataset Splits Yes The experiments have been conducted on the Movie Lens 10M data set using the original split with a fixed number of 200 iterations for all experiments.
Hardware Specification No The paper discusses runtime comparisons but does not specify the hardware (e.g., CPU, GPU models, memory) on which these experiments were conducted.
Software Dependencies No The paper mentions software like Cython, CXSparse, and scikit-learn, but does not provide specific version numbers for these dependencies, which are necessary for reproducibility.
Experiment Setup Yes fm = mcmc.FMClassification(init std=0.01, rank=8) fm = als.FMRegression(init std=0.01, rank=8, l2 reg=2) The experiments have been conducted on the Movie Lens 10M data set using the original split with a fixed number of 200 iterations for all experiments.