Welcome to the Quantum Realm. 

Enjoy today’s breakdown of news, research, & events within quantum.

Where error correction has dominated discussions of fault tolerance, Oxford Ionics challenges the status quo. Plus, Quantinuum's noble pursuit of science, quantum for the kids, and further evidence that HPC-QC is the future.

🗓️ UPCOMING

Sunday, July 14 | QTM-X Quantum Education Series 6 of 10: Quantum Hardware

📰 QUANTUM QUICK BYTES

💥 Oxford Ionics throws error correction to the wind: If told yesterday to envision a quantum chip that is powerful without extensive error correction AND can leverage existing semiconductor fabrication infrastructure, it would have sounded a lot like a scene from a fault-tolerant future. But, in a timeline that is very much this one, Oxford Ionics today announced their quantum chip, equipped with a patented Electronic Qubit Control system. With very minimal error correction, 10x fewer qubits, 99.97% two-qubit gate fidelity, and 99.9992% single-qubit operations, quantum boardrooms stood still today. Their innovative approach integrates the highest-performing qubit technology into a silicon chip, making high-performance quantum computing within reach through mass production and scalability.

🧪 Quantinuum’s part in Europe’s scientific pursuit: Quantinuum has just announced a partnership with the STFC Hartree Centre in a move that aligns with the quantum company’s commitment to scientific discovery. The Hartree Centre is part of the Science and Technology Facilities Council, one of Europe’s largest scientific research organizations. The Hartree Centre collaborates with businesses to assist them in adopting advanced technologies such as AI, data science, and supercomputing. The availability of on-site and cloud quantum computing through Quantinuum’s H-Series is sensible and powerful for the community the center serves. This collaboration will support advancements in quantum chemistry, computational biology, quantum AI, cybersecurity, and drive economic growth through scientific progress.

⁉️ The answer to the ultimate question of life, the universe, and everything is…34? What is the significance of 34 qubits? According to science, nothing newsworthy. Regardless, speculation continues around why several nations have imposed arbitrary limits on the export of quantum computers with 34 or more qubits. While scientists and experts worldwide are unsure what to make of the reasoning (or lack of), what is clear is that these restrictions are counterproductive to global collaboration and ultimately a hindrance to advancement that relies on open scrutiny and peer review. The specificity of the restrictions suggests concerns of misuse, such as breaking encryption, but without transparency around the decision-making process, these limits appear misguided.

🔭 On the shoulders of gravitational wave detection: LIGO, the Laser Interferometer Gravitational-Wave Observatory, was the inspiration behind the ideation of an effective, distributed quantum network. Current methods in quantum network construction face their share of limitations. Photons sent through fiber-optic cables quickly lose information (attenuation), while photons bounced to a satellite and then back to Earth maintain better integrity due to the vacuum of space but have limited transmission. Scientists from the University of Chicago Pritzker School of Molecular Engineering have formulated a method based on existing gravitational-wave infrastructure that combines the best of both worlds. Vacuum tubes, similar to LIGO but less pressurized, could effectively transmit photons in a cross-country network without attenuation by focusing them with lenses. While still theoretical, the team will test with a tabletop model before potential large-scale construction.

🤝 Community-powered meets industry-powered: Collaborations between quantum companies are business as usual around here. Usually. However, the team-up of Kipu Quantum and PlanQK is particularly powerful. While most quantum companies focus on either hardware or software, Kipu Quantum toes the line with hardware-specific algorithms for industries such as pharmaceuticals, chemicals, logistics, and finance. PlanQK, a community-focused platform with access to most quantum backends and extensive resources for all levels of expertise, has a growing ecosystem of over 100 organizations. This strategic acquisition, following Kipu Quantum's €11.4 million seed funding round, will continue to promote accessibility in quantum computing and accelerate the commercialization of Kipu's algorithms and services.

🏫 Cleveland Clinic quantum computer is fueling more than just research: The marvel of neuroplasticity in neuroscience lies in the brain's ability to absorb new information and integrate it into neural pathways. The marvel of children and young adults is their propensity towards plasticity. While alarms sound worldwide that quantum computing education is severely lacking and the number of jobs too vast as compared to available talent, the Cleveland Clinic is putting its IBM Quantum System One to good use. For the third year in a row, the clinic has partnered with the Cleveland Metropolitan School District to host a workshop for school-aged children to learn about AI and quantum computing. Workshop leaders consistently remark on the ability of participants to grasp the complex topics that often make quantum computing seem inaccessible.

☕️ FRESHLY BREWED RESEARCH

🖥️ ANALYZING MACHINE LEARNING PERFORMANCE IN A HYBRID QUANTUM COMPUTING AND HPC ENVIRONMENT

QUICK BYTE: The performance and limitations of a quantum convolutional neural network trained using transfer learning were evaluated across different hybrid HPC-QC processes to determine the advantages and constraints of our current hardware. While challenges with real quantum hardware exposed long queue times and other limitations, quantum simulators demonstrated computational speedups, especially in using use HPC resources with GPU-capable simulators for parallel training​.

PRE-REQS: 

• Transfer learning involves taking a pre-trained model typically trained on a large dataset and then fine-tuning it for a smaller, specific dataset.

• Convolutional neural networks are deep learning algorithms, commonly used for image recognition. Structurally, they consist of multiple layers, including convolutional layers that apply filters to input data to detect patterns, followed by pooling layers that reduce the dimensionality of the data, and then fully connected layers that out the final classification results.

SIGNIFICANCE: When it comes to the future of quantum computing, while quantum computers are being considered for use cases where they dominate most if not all of the computation process, it is logical to consider that quantum computers will more likely than not be injected for a subset of tasks. Considering the state of classical computation, many decades of research have been poured into high-performance computing, and today’s HPCs are on the Exascale — 1 quintillion operations per second. Due to their ability to handle vast datasets, machine learning tasks are commonly performed on HPCs.

As datasets continue to grow, quantum machine learning is emerging as a practical way to demonstrate speedup over classical computation. Quantum computers handle machine learning relatively well due to their capability for linear matrix calculations and weighted cost functions.

This study explored the performance of running hybrid HPC and QC programs across various methods, including simulations as well as the use of real quantum hardware on-premise or in the cloud, to determine the performance of such tasks against current hardware constraints. By integrating quantum simulators in a hybrid QML workflow, the researchers sought to evaluate the benefits as well as the limitations of this approach.

RESULTS: 

• Frontier GPUs achieved 56% and 77% speedups compared to Frontier’s CPU and a local system, respectively, with further improvements when scaling to larger datasets and multiple threads

• Quantum simulators like default.qubit can maintain high accuracy (about 94%) with reduced runtimes, though efficiency varied with different simulators and backend configurations

• Distributed training across multiple GPUs showed optimal execution times, with 8 GPUs on Frontier achieving the fastest runtimes and maintaining high accuracy, despite some challenges with GPU communication overheads

• Running QML on real quantum hardware was found to be impractical due to long queue times and computational limitations, making it challenging to complete ML algorithms in a reasonable timeframe​

• Their overall advice for achieving the best performance is to run jobs on systems that combine classical and quantum computational abilities and to use HPC resources with GPU-capable simulators for efficient parallel training​

HONORABLE RESEARCH MENTIONS: 

QTRL combines quantum neural networks with classical reinforcement learning to create more efficient reinforcement learning models. By using Quantum-Train to generate classical policy network parameters, QTRL addresses challenges such as data encoding and quantum hardware dependency during inference. Experimental results show that QTRL achieves comparable or superior performance to classical methods while using significantly fewer parameters. —> link to paper

A scalable improvement is found in implementing the generalized Toffoli gate using trapped-ion-based qutrits due to a reduced number of required two-qubit gates. This approach is validated experimentally, with the implementation of a three-qubit Grover's search algorithm showing a 10% increase in average performance. —> link to paper

A scalable parameterized quantum circuits classifier is used to improve multi-category classification tasks. SPQCC uses parallel execution of identical quantum circuits to optimize trainable parameters through cross-entropy loss minimization. As demonstrated using the MNIST dataset, classification tasks surpass classical methods with fewer parameters and efficient training. —> link to paper

A method for estimating the fidelity of quantum states that works with any measurement protocol is introduced. Through simulations and experiments with a trapped-ion quantum computer, the method demonstrates efficient and accurate fidelity estimation with fewer measurement outcomes. Additionally, it can be extended to estimate other quantum observables, making it a versatile tool for assessing quantum states and processes in different quantum computing applications. —> link to paper

UNTIL TOMORROW.