ModLaNets: Learning Generalisable Dynamics via Modularity and Physical Inductive Bias
Authors: Yupu Lu, Shijie Lin, Guanqi Chen, Jia Pan
ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We examine our framework for modelling doublependulum or three-body systems with small training datasets, where our models achieve the best data efficiency and accuracy performance compared with counterparts. and 5. Experiments and Table 2. Experiment settings and results for part 1 and 2 |
| Researcher Affiliation | Collaboration | 1Department of Computer Science, The University of Hong Kong, Hong Kong SAR, China 2Centre for Garment Production Limited (Trans GP), Hong Kong SAR, China. |
| Pseudocode | Yes | Algorithm 1: Mod La Net Framework |
| Open Source Code | No | The paper does not provide an explicit statement about releasing its source code or a link to a code repository for the described methodology. |
| Open Datasets | No | The paper describes generating its own dataset by integrating Lagrangian dynamics and using SciPy functions, but it does not provide access information (link, DOI, or formal citation) to a publicly available or open dataset. |
| Dataset Splits | No | The paper mentions 'training' and 'testing' results (Table 2) but does not explicitly specify dataset split percentages or absolute sample counts for training, validation, and testing sets. |
| Hardware Specification | Yes | Our model was built using Py Torch (Paszke et al., 2019), and experiments were conducted in Ubuntu 20.04 using single core i7@3.7GHz. |
| Software Dependencies | No | The paper mentions software like PyTorch, Autograd, and SciPy with their corresponding citations, but does not provide specific version numbers for these software components. |
| Experiment Setup | Yes | Hyperparameters for tuning are learning rate (10 4-10 1), training epoch (1-20K), and activation functions. and During training, each model outputs the acceleration given input state (q, q). Then for optimisation we use L2 loss: L = qˆ q 2 |