Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1]

MamKO: Mamba-based Koopman operator for modeling and predictive control

Authors: Zhaoyang Li, Minghao Han, Xunyuan Yin

ICLR 2025 | Venue PDF | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental The modeling and control performance of the proposed method is evaluated through experiments on benchmark time-invariant and time-varying systems. The experimental results demonstrate the superiority of the proposed approach. Additionally, we perform ablation experiments to test the effectiveness of individual components of Mam KO. ... 5 EXPERIMENTS In this section, we will evaluate the performance of the Mam KO model in terms of modeling and control.
Researcher Affiliation Academia Zhaoyang Li1, Minghao Han2, Xunyuan Yin1,2, 1 School of Chemistry, Chemical Engineering and Biotechnology, Nanyang Technological University 2 Nanyang Environment and Water Research Institute, Nanyang Technological University EMAIL; EMAIL; EMAIL
Pseudocode No The paper describes methods and optimization problems using mathematical formulations and diagrams (e.g., Figure 1), but does not include any explicitly labeled 'Pseudocode' or 'Algorithm' blocks or figures.
Open Source Code No The paper does not contain any explicit statements regarding the release of source code for the methodology described, nor does it provide any links to a code repository.
Open Datasets Yes Five benchmark systems are included to evaluate the modeling and control performance of the Mam KO. The Cart Pole balancing system is included as a benchmark system for the controller design task, which has been widely used in deep reinforcement learning (DRL) research (Lillicrap, 2015; Haarnoja et al., 2018). ... The experiments are set based on Open AI Gym (Brockman et al., 2016).
Dataset Splits Yes For each environment, trajectories of state and action samples are gathered, generating a training set of 36, 000 samples, a validation set of 4, 000 samples, and a test set of 4, 000 samples. ... For the water treatment system, trajectories of state and action samples are gathered, generating a training set of 20, 000 samples, a validation set of 2, 000 samples, and a test set of 2, 000 samples.
Hardware Specification Yes The online control implementation of the control methods is conducted on a computer equipped with an Intel Core i9-13900K CPU and 128 GB DDR4 RAM.
Software Dependencies No The paper mentions software tools like 'Casadi' and 'Interior Point Optimizer (IPOPT)' but does not provide specific version numbers for them or any other key software dependencies or libraries used in the implementation.
Experiment Setup Yes The methods are trained to predict state sequences over a 30-step horizon. Hyperparameters for each method are listed in Appendix D. ... Table 4: Hyperparameters of Mam KO ... Batch Size 256 Learning rate 1e-3 Prediction horizon H 30