Optimal Control Operator Perspective and a Neural Adaptive Spectral Method
Authors: Mingquan Feng, Zhijie Chen, Yixin Huang, Yizhou Liu, Junchi Yan
AAAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Experiments on synthetic environments and a real-world dataset verify the effectiveness and efficiency of our approach, including substantial speedup in running time, and high-quality in- and out-of-distribution generalization. We give experiments on synthetic and real-world environments to evaluate our instance-solution operator framework. |
| Researcher Affiliation | Academia | 1School of Electronic Information & Electrical Engineering, Shanghai Jiao Tong University, Shanghai, China 2Siebel School of Computing and Data Science, University of Illinois at Urbana-Champaign, Urbana, IL, USA 3School of Artificial Intelligence & Mo E Lab of AI, Shanghai Jiao Tong University, Shanghai, China EMAIL, EMAIL |
| Pseudocode | Yes | The data generation process is shown in Alg. 1 in Appendix. |
| Open Source Code | Yes | Code https://github.com/Feng Mingquan-sjtu/NASM |
| Open Datasets | Yes | We use the Pushing dataset (Yu et al. 2016), a challenging dataset consisting of noisy, real-world data produced by an ABB IRB 120 industrial robotic arm (Fig. 4, right part). |
| Dataset Splits | Yes | The training data are sampled from ID, while validation data and benchmark sets are both sampled from ID and OOD separately. [...] We set the size of the OOD fine-tuning dataset and #epochs to be 20% of that of the ID training dataset. |
| Hardware Specification | No | First, the comparison of the wall-clock running time is shown in Fig. 3a as well as in the third column of Tab. 2, which shows that the neural operator solvers are much faster than the classic solvers, although they both tested on CPU for fairness. |
| Software Dependencies | No | The batch size is min(10, 000, N), and the optimizer is Adam (Kingma and Ba 2014). |
| Experiment Setup | Yes | For all systems and all neural models, the loss is the mean squared error defined below, with a dataset D of N samples: L = 1 N P i,j D N(i)(tj) u i (tj) 2., the learning rate starts from 0.01, decaying every 1,000 epochs at a rate of 0.9. The batch size is min(10, 000, N), and the optimizer is Adam (Kingma and Ba 2014). |