Deep Rank-One Tensor Functional Factorization for Multi-Dimensional Data Recovery
Authors: Yanyi Li, Xi Zhang, Yisi Luo, Deyu Meng
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Extensive experiments on multi-dimensional data recovery problems, such as image and video inpainting, image denoising, and hyperspectral mixed noise removal, showcase the effectiveness of the proposed method. |
| Researcher Affiliation | Academia | 1Xi an Jiaotong University, Xi an, China 2Pengcheng Laboratory, Shenzhen, China 3Macau University of Science and Technology, Macao, China EMAIL, EMAIL, EMAIL, EMAIL |
| Pseudocode | No | The paper describes the optimization problem using ADMM and outlines the sub-problems and multiplier updates in mathematical equations, but it does not present these steps in a structured pseudocode or algorithm block. |
| Open Source Code | No | For other hyperparameters, please refer to the source code. The paper mentions source code but does not provide a direct link to a repository, state that code is released, or specify where it can be accessed. |
| Open Datasets | Yes | We use publicly available datasets. For image and video inpainting, we use three types of multi-dimensional data including color images (House, Tree, Jelly Beans)1, natural HSIs (Feathers, Flowers, Thread Spools)2, and videos (News, Carphone, Hall Monitor)3 as testing data. ... 1https://sipi.usc.edu/database/database.php 2https://www.cs.columbia.edu/CAVE/databases/multispectral/ 3http://trace.eas.asu.edu/yuv/ 4https://lesun.weebly.com/hyperspectral-data-set.html |
| Dataset Splits | No | The paper describes sampling rates for missing data and noise standard deviations for different tasks, but it does not provide specific train/test/validation splits for the datasets used. |
| Hardware Specification | No | The paper mentions "favorable computational and space efficiency" and provides "Time" metrics in tables, but it does not specify any particular hardware components (e.g., GPU models, CPU types, or memory) used for running the experiments. |
| Software Dependencies | No | The paper mentions using the "Adam optimizer" and activation functions like "Leaky Re LU" and "sine activation", but it does not provide specific version numbers for any software frameworks, libraries, or programming languages used (e.g., Python, PyTorch, TensorFlow versions). |
| Experiment Setup | Yes | For all tasks, the rank parameter r is set as int(β min(n1, n2)), where ni (i = 1, 2) are the spatial sizes of the observed data, and β is a hyperparameter that is set to 0.7, 0.3, 0.3 for image and video inpainting, image denoising, and hyperspectral mixed noise removal. The depth of the mode-3 deep transform k is set to 3. The network depth of the factor INRs is set to 3. The sine activation sin(ω0 ) is used as nonlinear activation because of its Lipschitz continuous property. For color image inpainting, ω0 is set to 1, otherwise it is set to 2. The trade-off parameters γ1, γ2 in the loss function are set to 4 10 5 and 4 10 4 respectively. In the hyperspectral mixed noise removal model, we set γ = 0.01 and µ = 0.04. For other hyperparameters, please refer to the source code. |