High Rank Matrix Completion With Side Information
Authors: Yugang Wang, Ehsan Elhamifar
AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | By extensive experiments on both synthetic and real data, and, in particular, by studying the problem of multi-label learning, we demonstrate that our method outperforms existing techniques in both low-rank and high-rank data regimes. |
| Researcher Affiliation | Academia | Yugang Wang,1 Ehsan Elhamifar2 1 School of Mathematical Sciences, University of Electronic Science and Technology of China 2 College of Computer and Information Science, Northeastern University |
| Pseudocode | Yes | Algorithm 1 Input: Incomplete data matrix Y , side information matrices Fr, Fc and set of indices of observed entries Ω. Output: Y , A, Q. 1: Set k = 0. Initialize matrices Q0, Y 0, A0, Z0, V 0, Γ0 1, Γ0 2, Γ0 3 as zero matrices. 2: Update Ak+1 using the equation (13) or (14). 3: Update Qk+1 using the equations (18). 4: Update Y k+1 using the equation (19). 5: Update Zk+1 using (20) and (21). 6: Update V k+1 using (20) and (21). 7: Update the multipliers Γk+1 1 , Γk+1 2 , Γk+1 3 using (22). 8: Set k = k + 1. While not converged, go to step 2. 9: return Y , A, Q. |
| Open Source Code | No | For Maxide, IMC and SIM, we used the publically available codes. We implemented dirty IMC, since the code was not available. The paper does not state that the authors' own code is publicly available. |
| Open Datasets | Yes | Multi-label learning for audio classification. The Birds dataset consists of 645 ten-second audio files recording the sounds of 19 different species of birds as well as the sounds of environments, such as wind or rain (Briggs et al. 2013). Multi-label learning for music classification. The CAL500 dataset is a collection of 502 Western popular songs from 502 unique artists... (Turnbull et al. 2008). Multi-label learning for image classification. The NUS-WIDE dataset contains more than 269,648 images... (Spyromitros-Xioufis et al. 2014). |
| Dataset Splits | No | Given a ground-truth data matrix, Y , we drop the values of δ fraction of the entries uniformly at random and change δ from 0.1 to 0.9. This describes how missing entries are created for evaluation, but does not specify train/validation splits for model training or hyperparameter tuning. |
| Hardware Specification | No | No specific hardware details (like GPU/CPU models, memory, or specific machine configurations) are mentioned for the experimental setup. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers for their implementation. It only mentions using 'publically available codes' for other methods. |
| Experiment Setup | Yes | In the experiments, for our proposed method, we set λ = 10, ρ = 2 102, μ1 = μ2 = 105, β = 102, τ1 = 10 5, τ2 = 10 2 . Experimentally, we observed robust performance with respect to the change of these parameters. |