Hybrid Regularization Improves Diffusion-based Inverse Problem Solving
Authors: Hongkun Dou, Zeyu Li, Jinyang Du, Lijun Yang, Wen Yao, Yue Deng
ICLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experimental evaluations on benchmark datasets, including FFHQ and Image Net, demonstrate that our proposed framework not only achieves highly competitive results compared to state-of-the-art methods but also offers significant reductions in wall-clock time and memory consumption. (Abstract) The entire Section 4: "EXPERIMENTS" details extensive empirical studies, evaluations on datasets, comparisons to baselines, and ablation studies, including quantitative results in tables and qualitative results in figures. |
| Researcher Affiliation | Academia | Hongkun Dou1, Zeyu Li1, Jinyang Du1, 3, Lijun Yang1, Wen Yao2, & Yue Deng1, 3, 1 Beihang University 2 Chinese Academy of Military Science 3 Beijing Zhongguancun Academy EMAIL, EMAIL |
| Pseudocode | Yes | Algorithm 1 Sampling procedure for HRDIS. (Page 5) and Listing 1: Pseudocode of HRDIS for performing one optimization step. (Page 12) |
| Open Source Code | Yes | Code is available at https://github.com/deng-ai-lab/HRDIS. (Abstract) |
| Open Datasets | Yes | Our experiments are conducted on the Image Net 256 256 (Deng et al., 2009) and FFHQ 256 256 (Karras et al., 2019) datasets, with results derived from 1k validation images... (Section 4) |
| Dataset Splits | Yes | Our experiments are conducted on the Image Net 256 256 (Deng et al., 2009) and FFHQ 256 256 (Karras et al., 2019) datasets, with results derived from 1k validation images, consistent with previous research standards (Chung et al., 2022; Mardani et al., 2024). We evaluate our results using two widely adopted metrics: Learned Perceptual Image Patch Similarity (LPIPS) (Zhang et al., 2018) and Fréchet Inception Distance (FID) (Heusel et al., 2017), computed on the test images. (Section 4) |
| Hardware Specification | Yes | All experiments, including time and memory calculations, were conducted on a single NVIDIA RTX 3090 GPU. (Appendix B.2) |
| Software Dependencies | No | Below, we give the Pytorch-style pseudocode for our HRDIS implementation. (Appendix B.2) The paper mentions "Pytorch-style pseudocode" but does not specify the version of PyTorch or any other key software libraries used with their versions, such as Python or CUDA. |
| Experiment Setup | Yes | We employed the Adam optimizer (Kingma & Ba, 2014) with momentum parameters set to (0.9, 0.99). The parameter β, utilized in synthesizing the hybrid noise, remained fixed at 0.2 throughout our experiments. We also chose descending timestep from t = 1 to t = 0 as in (Mardani et al., 2024). The denoiser weight ωt is determined as the inverse signal-to-noise ratio (SNR), σt/αt. In practice, we introduce two coefficients to balance the reconstruction and regularization terms, µLHR = λ1 µ y A(µ) 2 2 + λ2Et,ε,ϵ[ωt(ϵθ(µt, t) ϵhybrid)]. For most tasks, optimizing the number of steps in the range of N = 100 150 produces satisfactory results. For more challenging degenerations, such as compressed sensing and phase retrieval, we use more steps to improve performance further. In addition, for the phase retrieval task... Table 5 details the selected hyperparameters of our proposed method. (Appendix B.2 and Table 5) |