Clipped Matrix Completion: A Remedy for Ceiling Effects
Authors: Takeshi Teshima, Miao Xu, Issei Sato, Masashi Sugiyama5151-5158
AAAI 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We demonstrate the effectiveness of the proposed methods through experiments using both synthetic and benchmark data for recommendation systems. In this section, we show the results of experiments to compare the proposed CMC methods to the MC methods. |
| Researcher Affiliation | Academia | Takeshi Teshima,1,2 Miao Xu,2 Issei Sato,1,2 Masashi Sugiyama1,2 1The University of Tokyo 2RIKEN |
| Pseudocode | No | No explicit pseudocode or clearly labeled algorithm blocks were found in the paper. |
| Open Source Code | No | No explicit statement providing a link to open-source code for the methodology was found. A link to an arXiv preprint is provided, but this does not imply code availability. |
| Open Datasets | Yes | We used the following benchmark data sets of recommendation systems. Film Trust (Guo, Zhang, and Yorke-Smith 2013) consists of ratings obtained from 1,508 users to 2,071 movies on a scale from 0.5 to 4.0 with a stride of 0.5 (approximately 99.0% missing). For ease of comparison, we doubled the ratings so that they are integers from 1 to 8. Movielens (100K) consists of ratings obtained from 943 users to 1,682 movies on an integer scale from 1 to 5 (approximately 94.8% missing). |
| Dataset Splits | Yes | The generated elements of matrix M were randomly split into three parts with ratio (0.8, 0.1, 0.1). Then the first part was clipped at the threshold C (varied over {5, 6, 7, 8, 9, 11, 13}) to generate the training matrix Mc Ω(therefore, p = 0.8). The remaining two parts (without thresholding) were treated as the validation (Mv) and testing (Mt) matrices, respectively. In both experiments, we first split the observed entries into three groups with ratio (0.8, 0.1, 0.1), which were used as training, validation, and test entries. |
| Hardware Specification | No | No specific hardware details (like exact GPU/CPU models or detailed computer specifications) used for running experiments were provided. |
| Software Dependencies | No | No specific ancillary software details, such as library names with version numbers, were provided. |
| Experiment Setup | No | The paper mentions "hyperparameter tuning" and how they were selected ("by the f1 score on the validation entries"), but it does not provide concrete hyperparameter values (e.g., learning rates, batch sizes, number of epochs) or other system-level training configurations in the main text. |