Reinforcement Learning for Molecular Design Guided by Quantum Mechanics

Authors: Gregor Simm, Robert Pinsler, Jose Miguel Hernandez-Lobato

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We perform experiments to evaluate the performance of the policy introduced in Section 3. In our experiments, we show that our agent can efficiently learn to solve these tasks from scratch by working in a translation and rotation invariant state-action space.
Researcher Affiliation Academia 1Department of Engineering, University of Cambridge, Cambridge, UK.
Pseudocode No The paper describes the methodology in text but does not include structured pseudocode or algorithm blocks.
Open Source Code Yes Source code of the agent and environment is available at https://github.com/gncs/molgym.
Open Datasets Yes We train the agent on the multi-bag task using all formulas contained in the QM9 dataset (Ruddigkeit et al., 2012; Ramakrishnan et al., 2014) with up to 5 atoms, resulting in 11 bags (see Table 1).
Dataset Splits No The paper describes training an RL agent to generate molecules but does not specify explicit train/validation/test dataset splits with percentages or sample counts for the data used to train or evaluate the agent.
Hardware Specification Yes Experiments were run on a 16-core Intel Xeon Skylake 6142 CPU with 2.6GHz and 96GB RAM.
Software Dependencies Yes Structures are considered valid if they can be successfully parsed by RDKIT (Landrum, 2019).
Experiment Setup No Details on the model architecture and hyperparameters are in the Appendix.