Implementation of reinforcement learning in chemical reaction networks: application to phototaxis as curiosity-driven exploration

Short summary

Researchers frame unicellular algae navigation as a reinforcement learning problem, linking Partially Observable Markov Decision Processes with biochemical reaction dynamics. Using Inverse RL on experimental Chlamydomonas trajectories, they show how information-seeking emerges as an adaptive strategy. The model demonstrates that cellular run-tumble behavior implements active sensory sampling.

• •Novel framework linking POMDP with biochemical reaction networks for cellular navigation
• •Inverse RL infers behavioral objectives from experimental algae trajectories
• •Run-tumble behavior emerges as information-acquisition strategy for resolving sensory ambiguity

Generated with AI, which can make mistakes.