Dimension Reduction for Quantum Adaptive Agents
The paper introduces a 'route-truncate-repair' procedure to compress quantum adaptive agents. By routing the agent's dynamics through an input process, it maps the agent to a Matrix Product State (MPS) representation. This allows for dimension reduction in memory while maintaining certified accuracy guarantees, trading small fidelity loss for significant memory savings.