Quantum Control Methods for Cat Code Encoding Pulses: GRAPE & RL

Quantum Control Methods for Cat Code Encoding Pulses: GRAPE & RL

by Kunho Jang

Formal & Physical Sciences Physics PHPhysicsPHQQuantum physics (quantum mechanics and quantum field theory)
🎙 Kunho Jang 👥 267 📅 December 1, 2025 ⏱ 26 min 👁 70 📄 original study 🧭 2026-08-15
Available in: English (current) Français

Keywords

quantum optimal controlcat codeGRAPEreinforcement learningfidelity

Summary

The presentation by Kunho Jang, a high school student, introduces quantum optimal control methods for preparing cat code states in superconducting resonator systems. It begins with an overview of quantum error correction and cat codes, highlighting their hardware efficiency and robustness to certain noise channels. The need for high-fidelity state preparation motivates the use of quantum optimal control. The first method, GRAPE (Gradient Ascent Pulse Engineering), is a model-based numerical optimization that designs control pulses by maximizing fidelity. The presenter simulates six representative state transfers in a coupled resonator-transmon system, using the Q-Tip library, achieving high fidelities in both closed and open system simulations, with fidelities exceeding 99.4% even under noise. However, GRAPE’s reliance on an accurate system model is a limitation. To address this, the presenter introduces deep reinforcement learning, specifically Proximal Policy Optimization (PPO), as a model-free alternative. The talk covers the theoretical framework of RL, including Markov decision processes, policy, and the clipped objective function of PPO, which ensures stable updates. The presenter plans to implement PPO and compare its performance with GRAPE, possibly using GRAPE results for pre-training. The presentation concludes that both methods are promising for achieving fault tolerance in cat codes.

197 words

Critical Evaluation

Value of the Information & Strength of the Argument

The presentation provides a clear and structured introduction to two quantum control methods, with a focus on their theoretical foundations and practical implementation. The value lies in the demonstration of GRAPE simulations with realistic parameters and the discussion of RL as a model-free alternative. The argumentation is logical, moving from the problem of quantum error correction to the need for optimal control, and then comparing the two approaches. However, the depth is limited by the presenter’s level, and the RL part is mostly theoretical without experimental results.

Scientific Rigor, Source Quality, Title Accuracy

The presentation cites two key papers: ‘Encoding of qubit states in resonators with cat codes’ by Winther and ‘Model-free quantum control with reinforcement learning’ by VVAC et al. These are relevant and credible sources. The title accurately reflects the content. The methodology is described with sufficient detail for reproducibility, including parameters and simulation setup. However, the lack of peer review and the presenter’s background limit the overall rigor.

170 words

Title / Content Match

The title accurately reflects the content, focusing on GRAPE and reinforcement learning for cat code encoding pulses.

Quality & Reliability

7/10

The presentation is based on established quantum control methods and includes original simulation results. However, it is a high school student's seminar, so depth and peer-review are limited.

Key Moments

Markers derived by PSI from the transcript: the creator did not define chapters.

Cited Sources

  • Encoding of qubit states in resonators with cat codes — Referenced as the paper for cat code encoding.
  • Model-free quantum control with reinforcement learning — Referenced as the paper for RL-based quantum control.

Concurring Sources

Contribution & Novelties

The presentation offers a comparative overview of GRAPE and PPO for cat code state preparation, with original simulation results for GRAPE. It highlights the model-based vs. model-free trade-off and suggests future directions like pre-training PPO with GRAPE pulses. The novelty is limited as it is a seminar presentation, but it provides a clear educational synthesis.

Pour aller plus loin :

82 words

Radar Profile

The radar profile shows balanced scores across information quantity, quality, technical level, and reliability, indicating a well-rounded presentation. The slightly lower reliability reflects the lack of peer review and the presenter's student status.

Reliability 7/10