Coherent Ising Computing Platform Achieves 55% Success Rate on 100-Vertex Mobius Ladder Graphs
Summarize this article with:
Coherent Ising Machines represent a promising new approach to tackling complex computational problems, and a team led by Hai Wei, Chengjun Ai, and Putuo Guo from Beijing QBoson Quantum Technology Co., Ltd., has now demonstrated a versatile platform with significantly improved performance.
This research overcomes key limitations of earlier CIM designs, notably susceptibility to errors, and achieves a remarkable 55% success rate in identifying optimal solutions for challenging problems involving 100 variables.
The team’s innovative use of femtosecond laser technology delivers higher peak power with lower energy consumption, and importantly, maintains a consistent success rate over an extended eight-hour period, highlighting the potential for practical applications in fields ranging from molecular docking to financial credit scoring. This achievement substantiates the theoretical promise of coherent Ising computing and establishes a robust foundation for developing larger, more powerful systems.
Femtosecond Lasers Drive Chip-Scale Ising Machine This research details the development of a fast, chip-scale Ising Machine (CIM) powered by femtosecond (fs) lasers. This CIM aims to solve complex optimization problems more quickly than traditional methods and potentially outperform other CIM implementations. The core of this technology lies in its ability to rapidly switch optical paths within a compact device using incredibly short laser pulses, opening possibilities for wider deployment of this computational approach. The researchers demonstrated a significant speed advantage over conventional Simulated Annealing and a previous CIM design when solving benchmark problems, such as the Möbius Ladder graph. This advantage becomes more pronounced as the complexity of the graph increases. Crucially, the team implemented temperature control to stabilize the optical paths, ensuring reliable and consistent computation. The CIM successfully tackled three challenging optimization problems: the Möbius Ladder graph, random graphs, and even molecular docking and credit scoring. When applied to the Möbius Ladder graph, the CIM achieved a 100% success rate for 20-vertex graphs and maintained high success rates for larger graphs, significantly outperforming existing methods. In molecular docking, the CIM accurately predicted the structure of molecules, achieving low root-mean-square deviation values compared to known crystal structures, suggesting potential applications in drug discovery. Furthermore, the CIM improved the performance of a machine learning model used for credit scoring by effectively selecting the most relevant features. These results demonstrate the potential of this CIM for real-world applications in diverse fields, including finance, materials science, and drug discovery. This work represents a significant step forward in CIM technology, paving the way for faster and more efficient computation of complex optimization problems. Future research will focus on improving device stability, increasing scalability, and developing algorithms specifically tailored to this unique CIM architecture. This innovative approach to building CIMs using femtosecond laser technology promises to unlock new possibilities in computational problem-solving.
Femtosecond Laser Enhanced DOPO Spin Preparation Researchers engineered a system utilizing femtosecond (fs) laser pulses to enhance non-linearity during the preparation of pulses within a Degenerate Optical Parametric Oscillator (DOPO), which represent the spins within a Coherent Ising Machine (CIM). This technique, coupled with a newly developed cavity stabilization system, enables DOPO generation at lower average power, amplifying quantum effects and ultimately improving the accuracy of solutions. The experimental setup centers around a laser delivering pulses lasting approximately 100 femto-seconds with a repetition rate of 100 million pulses per second, providing excellent timing stability. A process called Second Harmonic Generation creates pulses of a different wavelength, which are then used to generate the DOPO pulses representing the spins within the CIM. The system carefully aligns these pulses within the cavity, ensuring precise synchronization. This innovative system achieved a 55% success rate in identifying optimal solutions for a complex problem, representing the highest reported success rate among CIMs and other emerging quantum technologies. Furthermore, the CIM demonstrated robust stability, maintaining this performance level for over eight hours of continuous operation. A cloud platform was established to provide access to users across various industries, facilitating wider adoption and testing. Experiments revealed an average success rate of 55% in identifying optimal solutions for a complex problem, representing the highest reported success rate among CIMs and other emerging quantum technologies.
The team employed a laser delivering pulses lasting approximately 100 femto-seconds with a repetition rate of 100 million pulses per second, providing excellent timing stability. This laser source, coupled with a process called Second Harmonic Generation, produced pulses crucial for generating the DOPO pulses representing the spins within the CIM. Measurements confirm that this femtosecond pulse approach results in higher peak power and allows for operation at lower power, leading to more pronounced quantum effects. Furthermore, the CIM system demonstrated remarkable operational stability, maintaining the 55% success rate for a continuous period of eight hours. This sustained performance underscores the practical applicability of CIMs for real-world scenarios and facilitated the establishment of a cloud platform to host user requests. Researchers successfully applied the system to molecular docking and credit scoring tasks, illustrating its potential across diverse industries. These results substantiate the theoretical promise of CIMs and pave the way for their integration into large-scale practical applications. Stable Optimization with a Fiber Ising Machine Researchers have demonstrated a coherent Ising machine capable of solving complex optimization problems with improved accuracy and stability.
The team achieved a 55% success rate in identifying optimal solutions within a complex problem, a significant advancement in the field. This was accomplished through the implementation of femtosecond laser pumping, which resulted in higher peak power and reduced power requirements for fiber-based systems. Notably, the system maintained stable calculations for over eight hours, highlighting its potential for practical applications. The researchers further validated the machine’s capabilities by successfully applying it to problems in molecular docking and credit scoring, demonstrating its versatility across different optimization areas. While acknowledging current limitations in scaling the system, specifically challenges related to laser repetition rates and fiber length, the team suggests future work could focus on increasing these parameters. They also note that while progress has been made in integrating components onto chips for improved stability, fabricating kilometer-scale on-chip fiber remains a hurdle. These findings encourage further exploration of coherent Ising machines as a viable alternative technology for tackling complex optimization challenges in the future. 👉 More information 🗞 A versatile coherent Ising computing platform 🧠 ArXiv: https://arxiv.org/abs/2512.07182 Tags: