Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
FRN: Fractal-Based Recursive Spectral Reconstruction Network
Authors: Ge Meng, Zhongnan Cai, Ruizhe Chen, Jingyan Tu, Yingying Wang, Yue Huang, Xinghao Ding
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Through extensive experimentation across different datasets, FRN achieves superior reconstruction performance compared to state-of-the-art methods. 4 Empirical Results |
| Researcher Affiliation | Academia | Ge Meng1, Zhongnan Cai1, Ruizhe Chen1, Jingyan Tu1, Yingying Wang1, Yue Huang1, Xinghao Ding1, 1Key Laboratory of Multimedia Trusted Perception and Efficient Computing, Ministry of Education of China, Xiamen University, Xiamen, Fujian, China EMAIL |
| Pseudocode | No | The paper describes the method using text and figures (Figure 2, Figure 3) but does not include a dedicated pseudocode or algorithm block. |
| Open Source Code | Yes | Code is available at https://github.com/mongko007/frn. |
| Open Datasets | Yes | Dataset. To validate the effectiveness of the proposed network, we conducted experiments on two datasets. The first dataset is the CAVE dataset [54] provided by Columbia University... The second dataset is the Harvard dataset [14] provided by Harvard University. |
| Dataset Splits | Yes | We randomly selected 20 HSIs for training, 6 HSIs for validation, and 6 HSIs for testing. The second dataset is the Harvard dataset [14] provided by Harvard University. It contains 50 HSIs... We randomly selected 30 HSIs for training, 10 HSIs for validation, and 10 HSIs for testing. |
| Hardware Specification | Yes | We implemented our network on the PC with a single NVIDIA RTX 4090 GPU and built it in the Py Torch framework. |
| Software Dependencies | No | We implemented our network on the PC with a single NVIDIA RTX 4090 GPU and built it in the Py Torch framework. |
| Experiment Setup | Yes | In the training phase, the Adam optimizer [17] was used to optimize the model parameters. The initial learning rate was set to 4 10 4 , and the learning rate was decayed using a cosine annealing schedule with a minimum value of 1 10 6. The batch size was set to 32. We cropped 64 64 patches from 3D cubes and input them into the network. We set the number of recursive levels to 5, where each atomic generation module reconstructs a two-channel image. The threshold parameter α in Eq. (8) is empirically set to 0.5. |