DeepUnfolding_Agent: A Classical-Quantum Adaptive Control API
Authors: Shanika Iroshi Nanayakkara, Shiva Raj Pokhrel
Subjects: Computation and Language (cs.CL); Machine Learning (cs.LG)
Submitted: 21 July 2026
arXiv: 2610.20615 [cs.CL]
DOI: 10.48550/arXiv.2610.20615
Overview
Adaptive optimization workflows in machine learning, federated learning, quantum machine learning, and hybrid classical-quantum systems frequently embed controller logic directly inside task-specific training scripts. This tight coupling limits reuse, reproducibility, safety validation, and fair comparison across execution environments.
This paper presents DeepUnfolding_Agent, a schema-driven FastAPI platform that exposes deep-unfolded adaptive control as an external software service. External executors register a strategy schema, submit round-level diagnostic variables, receive schema-bounded control actions, execute one optimization round, and return feedback for controller updates.
Architecture
The framework separates controller intelligence from domain-specific execution through a modular pipeline:
- Strategy registry — registers and manages controller strategies per domain
- Diagnostic-to-state adapter — normalizes incoming round-level diagnostics into controller-consumable state
- Controller factory — instantiates deep-unfolded controllers on demand
- Safety guard — constrains emitted actions within schema-defined bounds
- Executor abstraction — decouples the control API from the calling execution environment
- Feedback update path — closes the loop with returned round feedback
- SQLite persistence layer — provides traceable state and checkpointing
- Docker deployment — ensures portability across runtimes
- Reproducible benchmark scripts — support fair, repeatable comparison
Execution Pathways
The implementation supports multiple pathways:
- Classical
- Quantum-aware
- Hybrid classical-quantum
- Qiskit Aer simulation
- Optional IBM Quantum processing (hardware-backed)
Validation
The software is validated through unit, schema, API, safety, orchestration, quantum execution, checkpointing, and Docker tests. External user workflows are demonstrated across fixed versus adaptive control regimes in:
- Finance
- Healthcare
- Genomics
- Transport
- Noisy quantum simulation
- Hardware-backed IBM Quantum execution
Results
The results show that DeepUnfolding_Agent provides a reusable and traceable adaptive control layer for heterogeneous optimization workflows. The quantum experiments are interpreted strictly as API-level quantum control and hardware portability evidence, rather than as claims of quantum advantage.
Key Contributions
- A controller-as-a-service abstraction that decouples adaptive control logic from task-specific training code.
- A schema-driven interface that enables safety validation and fair cross-environment comparison.
- A modular, testable architecture with persistence, containerization, and reproducible benchmarks.
- Demonstrated portability across classical, simulated quantum, and hardware-backed quantum execution pathways.
Citation
Nanayakkara, S. I., & Pokhrel, S. R. (2026).
DeepUnfolding_Agent: A Classical_Quantum Adaptive Control API.
arXiv:2610.20615 [cs.CL].
https://doi.org/10.48550/arXiv.2610.20615
via ArXiv CL+LG