Welcome to the Quantum Realm.

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

🧠 A neuron model from the early 1900s is the inspiration behind the first demonstration of a hybrid quantum/neuromorphic QSNN and QSCNN. Plus, Moody’s and QuEra are using reservoir computing to tackle tropical cyclone intensity predictions, University of Chicago researchers have developed a chip that could scale the connection of quantum systems, and Quantum Transistors is working towards a universal QPU.

🗓️UPCOMING

Sunday, July 28 | QTM-X Quantum Education Series 7 of 10

📰QUANTUM QUICK BYTES

🌪️ Moody’s and QuEra are using quantum computing for tropical cyclone intensity predictions: Moody’s is experimenting with quantum computing to improve the accuracy of tropical cyclone predictions and improve their risk management services suite. Partnering with QuEra, they plan to apply reservoir computing to address biases in meteorological data used for forecasting cyclone paths and intensities. Additionally, Moody’s is investigating other quantum computing use cases, including financial forecasting and portfolio optimization, to understand the potential benefits and applications of quantum technology in their risk assessment services.

🧱 The EQUALITY project is developing a comprehensive stack of quantum computing technologies: A consortium, comprised of prominent partners like Airbus and Capgemini and funded by Horizon Europe, is on a mission to address computational challenges in aerospace, energy, battery design, material development, and space optimization through the EQUALITY project. The project is focused on creating a full stack of quantum computing technologies, including core algorithms, hardware, middleware, and APIs, to fully realize the potential of NISQ hardware for industrial applications. Over the first 18 months, the consortium has made advancements in areas such as optimization, noise analysis, and circuit tailoring for efficient use of quantum resources. Notable developments include approaches to circuit cutting, analog solutions for partial differential equations, quantum noise estimation, hardware-oriented circuit compilation, and an API for quantum routines addressing partial differential equations. Overall, the EQUALITY project is on track to deliver a competitive edge to European industries.

⚡️ Quantum Transistors has secured funding from the European Innovation Council to scale quantum processor development: Quantum Transistors has been awarded up to €17.5 million by the European Innovation Council to advance its quantum processor technology. The funding will support the company's R&D expansion and team growth. CEO Shmuel Bachinsky emphasized that this investment validates their vision to democratize quantum computing, much like the Intel 8086 did for classical computing. Quantum Transistors is developing a universal quantum processor on a single chip by using native photonics to reduce inter-qubit noise and integrate into standard data centers.

🤝 Singapore’s National Quantum Office and Quantinuum have signed an MoU to collaborate on quantum computing application: Singapore’s National Quantum Office, A*STAR, NUS, NSCC, and Quantinuum have recently signed a MoU to use Quantinuum’s advanced H-Series and Helios quantum computers for joint research and development in quantum computing applications, with a focus on computational biology. Singaporean scientists will be able to use Quantinuum’s quantum processors for modeling complex biological systems, advancing drug discovery, and advancing personalized medicine. Quantinuum, in turn, plans to establish a dedicated R&D presence in Singapore to encourage knowledge exchange as well as further development of quantum applications and algorithms.

🔗 Researchers at the University of Chicago have developed a semi-open chip design to integrate atom arrays with photonic circuits: Scaling quantum information systems will require connecting multiple quantum systems. Researchers at the University of Chicago have developed a semi-open chip design that allows atom arrays trapped in optical tweezers to interact seamlessly with photonic circuits. This addresses the challenge of integrating photonic devices, which can disrupt delicate quantum states. The chip design features separate regions for computation and connection which allows for efficient quantum computations and data transfer via optical fibers. This setup could be one of the first steps along the path to creating larger and more interconnected quantum computing platforms.

☕️FRESHLY BREWED RESEARCH

A QUANTUM LEAKY INTEGRATE-AND-FIRE SPIKING NEURON AND NETWORK

QUICK BYTE: A quantum leaky integrate-and-fire neuron is used as the basis for quantum spiking neural networks and quantum spiking convolutional neural networks. These models demonstrate competitive accuracy and energy efficiency compared to classical and quantum counterparts, as well as serve as the first demonstration of such networks.

PRE-REQS:

• The synapse is the junction between two neurons where neurotransmitters are released to transmit signals.

• The membrane potential is the electrical potential difference across a neuron’s cell membrane.

SIGNIFICANCE: The ingenuity of AI lies in its design, which is inspired by the organic brain. Artificial neural network neurons are designed to mimic the synapse and node interactions of neurons in our brain, however, as an abstraction of the original design. While ANNs have been the most widely used and successful in computation up until now, they are not the only neural network model.

Another neuron model, proposed in 1907, is the LIF neuron — leaky integrate-and-fire neuron. This neuron models how biological neurons integrate inputs through synapses and generate outputs through changes in the membrane potential and firing spikes (action potentials). The LIF model can be decomposed as follows:

• Integration - the neuron integrates synaptic inputs and incoming spikes increase the membrane potential

• Leaky - the membrane’s potential decays over time; natural tendency to return to its resting state

• Fire - the neuron fires an action potential when the membrane potential reaches a certain threshold; returns the membrane potential to a value lower than the resting state

This neuron model has been proposed for use in neuromorphic computing neural networks, such as spiking neural networks. The advantage of SNNs as compared to ANNs is that they encode temporal data whereas ANNs do not. So, SNNs can be considered as more data-rich while still exceeding ANNs in energy efficiency.

Where ANNs have stood the test of time, which should arguably not be taken lightly, we are running into difficulties related to power consumption which have led many to consider alternative methods of computation. Quantum computing has the potential to outperform ANNs in specific tasks as well as reduce energy consumption, but NISQ hardware struggles to compete with ANN architecture.

The authors propose a hybrid quantum and neuromorphic computing neural network using quantum leaky integrate-and-fire neurons to address the problem of energy efficiency while taking advantage of quantum’s ability to handle high-dimensional and complex datasets.

They develop, for the first time, QSCNN (quantum spiking convolutional neural network) and QSNN (quantum spiking neural network) using QLIF neurons to apply to the MNIST, Fashion-MNIST, and Kuzushiji-MNIST datasets, alongside their fully classical and quantum counterparts to evaluate accuracy and efficiency across models. All models were trained using a training set of 60K images, a test set of 10K images, and the same computer including both noise and noiseless quantum simulations to ensure consistency in results.

In the resulting comparison, the QSCNN and the QSNN are neither the most efficient nor the most accurate — however, they achieve competitive accuracy at a comparative speed to ANNs and improved speed compared to purely quantum models. The biggest takeaways here would then be using the LIF neuron as a closer representation of a biological neuron and greater energy efficiency. Also, while it did not surpass the ANN, this is the first demonstration of such a network and improvements are expected after iterations and refinement.

RESULTS:

• Introduced a quantum leaky integrate-and-fire neuron and implemented as a compact high-fidelity quantum circuit requiring only two rotation gates and no CNOT gates; used to construct a QSNN and QSCNN

• Demonstrated competitive accuracy compared to classical and quantum models; quantitatively, the QSNN and QSCNN were 24 and 68 times faster in noiseless simulations, and 146 and 333 times faster in noisy simulations, respectively

• QLIF-based networks offered improved energy efficiency over classical ANNs and quantum variational classifiers

• Proof-of-concept demonstration of QSNN using QLIF neurons and potential for hybrid quantum/neuromorphic computing to improve on both energy-efficiency and performance

HONORABLE RESEARCH MENTIONS:

This is the exploration of quantum computer architectures where interactions are mediated between hot qubits not in their mechanical ground state. It introduces a method to use logical qubit encoding to make gates resilient against thermal noise and position fluctuations. By enlarging the logical system, gate fidelities can be improved, providing a platform-independent tool to mitigate thermal noise in trapped-particle-based architectures. —> link to Quantum computation with logical gates between hot systems

A quantum-classical hybrid pipeline is designed to address real-world drug discovery challenges. The approach uses quantum computation for determining Gibbs free energy profiles for prodrug activation and simulating covalent bond interactions, with a focus on KRAS protein interactions with inhibitors. —> link to A hybrid quantum computing pipeline for real world drug discovery

A VQA for solving partial differential equations is introduced; uses Lagrange polynomial encoding combined with derivative quantum circuits and Hadamard test differentiation. The algorithm's effectiveness is demonstrated by solving the damped mass-spring system and the Poisson equation under various boundary conditions, ultimately showing reduced gate complexity and improved solution quality compared to existing VQAs. —> link to A New Variational Quantum Algorithm Based on Lagrange Polynomial Encoding to Solve Partial Differential Equations

UNTIL TOMORROW.