Alias-Free Mamba Neural Operator
Authors: Jianwei Zheng, Wei Li, Ni Xu, Junwei Zhu, XiaoxuLin , Xiaoqin Zhang
NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Mamba NOs are evaluated on a diverse set of benchmarks with possibly multi-scale solutions and set new state-of-the-art scores, yet with fewer parameters and better efficiency. |
| Researcher Affiliation | Academia | Jianwei Zheng, Wei Li, Ni Xu, Junwei Zhu, Xiaoxu Lin, and Xiaoqin Zhang Zhejiang University of Technology, Hangzhou, Zhejiang |
| Pseudocode | Yes | Algorithm 1 SSM Block (Mamba Integration) |
| Open Source Code | Yes | Codes are available 2. https://github.com/ZhengJianwei2/Mamba-Neural-Operator. |
| Open Datasets | Yes | Representative PDE Benchmarks (RPB). As a standard set of benchmarks for machine learning of PDEs, RBP focuses solely on two-dimensional PDEs... on a larger dataset[33] (10,000 samples, with 7,000 used for training) |
| Dataset Splits | Yes | For the training phase, we produce 1024 samples, and for the evaluation phase, we furnish 256 samples each for both in-distribution and out-of-distribution testing... A validation set, composed of 128 samples, is also constructed for model selection purpose. |
| Hardware Specification | Yes | For fairness and reliability, all experiments are consistently conducted on standardized platform with an NVIDIA RTX 3090 GPU and 2.40GHz Intel(R) Xeon(R) Silver 4210R CPU. |
| Software Dependencies | No | The paper mentions general deep learning frameworks and tools but does not provide specific version numbers for software dependencies such as PyTorch, TensorFlow, CUDA, or other libraries. |
| Experiment Setup | Yes | Training Details and Baselines. For fairness and reliability, all experiments are consistently conducted on standardized platform... Table 8: Mamba NO best-performing hyperparameters configuration for different benchmarks. η, γ, and ω represent the learning rate, scheduler gamma, and weight decay, respectively. Convolution Integration and Mamba Integration denote the quantities of each in a single layer of kernel integration. |