Reinforcement Symbolic Regression Machine

Authors: Yilong Xu, Yang Liu, Hao Sun

ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We test the performance of our method on multiple different datasets and compare it with the following baseline models in symbolic learning:
Researcher Affiliation Academia Yilong Xu1, Yang Liu2, Hao Sun1, 1Gaoling School of Artificial Intelligence, Renmin University of China, Beijing, China; 2School of Engineering Science, University of Chinese Academy of Sciences, Beijing, China;
Pseudocode Yes Algorithm 1 Expression generation by RSRM
Open Source Code Yes Code and models of Reinforcement Symbolic Regression Machine(RSRM) are available at https://github.com/intell-sci-comput/RSRM.
Open Datasets Yes To evaluate the efficiency of our model, we first utilize four basic benchmark datasets (see Appendix C for details): Nguyen (Uy et al., 2011), Nguyenc (Mc Dermott et al., 2012), R (Mundhenk et al., 2021b), Livermore (Mundhenk et al., 2021b), and AIFeynman (Udrescu & Tegmark, 2020).
Dataset Splits Yes Each dataset is divided into three subsets: a training set, a test set, and a validation set. The training set comprises points ranging from 30 to 80, while the test set consists of points ranging from 10 to 25. The validation set covers a broader range, spanning from 0 to 100.
Hardware Specification No No specific hardware details (such as GPU/CPU models, processors, or memory) used for running the experiments were provided.
Software Dependencies No We treat each placeholder as an unknown variable, which is optimized to maximize the reward. The BFGS (Roger Fletcher & Sons, 2013) algorithm, available in the scipy (Virtanen et al., 2020) module in Python, is used for optimization. No specific version numbers for Python, SciPy, or DEAP are provided.
Experiment Setup Yes The full set of hyperparameters can be seen in Table S1.