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 and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in [1].
D-CODE: Discovering Closed-form ODEs from Observed Trajectories
Authors: Zhaozhi Qian, Krzysztof Kacprzyk, Mihaela van der Schaar
ICLR 2022 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In this section, we perform a series of simulations to evaluate whether the algorithms can discover the underlying closed-form ODEs that govern the observed trajectories. |
| Researcher Affiliation | Academia | Zhaozhi Qian University of Cambridge EMAIL Krzysztof Kacprzyk University of Cambridge EMAIL Mihaela van der Schaar University of Cambridge, UCLA, The Alan Turing Institute EMAIL |
| Pseudocode | Yes | The Pseudocode is presented in Algorithm 1. |
| Open Source Code | Yes | 2The code is available at https://github.com/Zhaozhi QIAN/D-CODE-ICLR-2022. |
| Open Datasets | Yes | We used the dataset collected by Wilkerson et al. (2017) based on eight clinical trials on cancer patients. |
| Dataset Splits | Yes | After data cleaning, we obtained 310 trajectories, among which 200 is used for training and 110 for evaluation. |
| Hardware Specification | Yes | Experiments are run on a computer with Intel Xeon E3-12xx v2 CPU (16 cores) and 60 GB memory. |
| Software Dependencies | No | The paper mentions software like gplearn, derivative, and sympy, along with citations. However, it does not provide specific version numbers for these software components. For example, it states 'We use the python package derivative for numerical differentiation (Quade & Goldschmidt, 2020)' but not a specific version like 'derivative 0.2.1'. |
| Experiment Setup | Yes | The hyperparameters of genetic programming is decided based on a pilot study... The values are listed below. ... 1. population size: 15000 2. tournament size: 20 3. p crossover: 0.6903 4. p subtree mutation: 0.133 5. p hoist mutation: 0.0361 6. p point mutation: 0.0905 7. generations: 20 8. parsimony coefficient: 0.01 |