Research map / Reinforcement learning / Visual and representation RL
Dynamic Control of Stochastic Evolution: A Deep Reinforcement Learning Approach to Adaptively Targeting Emergent Drug Resistance
Dalit Engelhardt
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train with 0.3 per-episode probability of mutation instead of mutations allowed in every episode with lower or higher per-step rate (ablation) | CelluDose DDPG training · training performance (drug resistance simulation) | training performance: P_epis,mut=0.3 vs P_epis,mut=1 with P_step,mut=0.3 times Phi | Improved |
| add the model-derived instantaneous growth rate as a state feature and shaping potential | CelluDose DDPG · training convergence | policy learning | Improved |
| add 20-fold lower mid-episode dosage penalty when population exceeds initial population (binary penalty function) (ablation) | CelluDose reward · training stability | stability and sensitivity to penalty weights: 20-fold lower weight | Improved |
| add positive bias initialization of the actor/critic from deterministic dose estimate (ablation) | CelluDose networks · training | avoids zero-dose local optimum | Improved |
| add an extra hidden layer to the actor network (ablation) | DDPG actor · training | performance | Improved |
| rescale cell concentrations by log(x+1)/log(K) (ablation) | CelluDose state features · training | convergence smoothness | Improved |
| raise mutation probability above the training value of 1e-6 up to 1e-3 for dual-drug policy | CelluDose dual-drug policy · mutation-rate robustness | success rate: 100% success up to P_step,mut=10^-3 | No clear effect |
| raise mutation probability to 1e-2 for dual-drug policy | CelluDose dual-drug policy · mutation-rate robustness | success rate: 100% -> 47% success at P_step,mut=10^-2 | Hurt |
| raise mutation probability up to 1e-1 for single-drug policy | CelluDose single-drug policy · mutation-rate robustness | success rate: 100% success up to P_step,mut=10^-1 | No clear effect |
| use wider hidden layers and more samples for dual-drug instead of single-drug training | CelluDose · dual-drug training | episodes to convergence: 50,000 -> 130,000 episodes | Mixed |
Closest papers by meaning
- Reinforcement Learning for Control of Non-Markovian Cellular Population Dynamics (ICLR 2025 · similarity 0.91)
- Uncertainty-Aware Multi-Objective Reinforcement Learning-Guided Diffusion Models for 3D De Novo Molecular Design (NeurIPS 2025 · similarity 0.82)
- Predicting Dose-Response Curves with Deep Neural Networks (ICML 2024 · similarity 0.81)
- Conservative Continuous-Time Treatment Optimization (NeurIPS 2026 · similarity 0.81)
- Stochastic Control for Fine-tuning Diffusion Models: Optimality, Regularity, and Convergence (ICML 2025 · similarity 0.81)
- Safe Exploration in Dose Finding Clinical Trials with Heterogeneous Participants (ICML 2024 · similarity 0.81)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
