Monday, September 30th, 2024

Enjoy a nice cup of freshly brewed quantum news ☕️ 

Today’s issue includes:

• Independent researchers explore the potential of quantum machine learning to classify ten types of cancer and identify novel biomarkers from high-dimensional RNA-seq datasets.

• Électricité de France (EDF) is collaborating with Pasqal to explore the potential of quantum computing for improving energy forecasting, material simulation, and smart charging.

• A subspace-preserving quantum convolutional neural network architecture uses Hamming-weight preserving quantum circuits for efficient image classification tasks, demonstrating polynomial speed-ups.

• Plus QubiCSV for more efficient data storage & management, carbo nano-onions, and QuanDOOM.

And even more research, news, & events within quantum.

⚕️ Quantum Machine Learning Used for Detection of Cancer Biomarkers, Outperforms Classical Models in Classifying High-Dimensional Data

QUICK BYTE: Independent researchers explore the potential of quantum machine learning to classify ten types of cancer and identify novel biomarkers from high-dimensional RNA-seq datasets, demonstrating the advantages of QML in handling complex biological data compared to classical machine learning models.

DETAILS

• Quantum support vector machine successfully classified cancer types and identified novel biomarkers, with some outperforming classical models such as support vector machine, random forest, and logistic regression. QSVM excelled in accuracy and precision specifically for cancers such as head and neck squamous cell carcinoma and kidney renal clear cell carcinoma, with overall performance better in high-dimensional, non-linear datasets.

• The QSVM model was simulated on classical computers using the QML scikit-qulacs package. The study used a hybrid classical-quantum approach, combining quantum feature mapping with Fisher scores for feature selection. Key technical steps included dimensionality reduction, oversampling using SMOTE, and parameterized quantum circuit construction, followed by performance analysis on test datasets.

• The study identified novel cancer biomarkers such as CMTM3 for bladder cancer and CDH3 for colon adenocarcinoma. DNA methylation analysis provided insight into the biological significance of gene expression variations between cancerous and normal tissues.

• This research highlights the potential of QML in biomedical applications, especially in the efficient handling of complex, high-dimensional datasets for cancer classification. QML’s place in medical research may service such applications as disease diagnostics and personalized treatment plans.

⚡️ EDF and Pasqal Take Advantage of Quantum Computing for Renewable Energy Forecasting and Smart Charging Challenges

QUICK BYTE: Électricité de France (EDF) is collaborating with Pasqal to explore the potential of quantum computing for improving energy forecasting, material simulation, and smart charging, in order to to overcome challenges related to renewable energy unpredictability.

DETAILS

• EDF faces challenges in accurately forecasting energy demand, specifically due to the unpredictable nature of renewable energy sources such as wind and solar power, and the complexities of managing energy distribution with nuclear and gas plants, which cannot quickly adjust to fluctuations in supply and demand.

• EDF began exploring quantum computing in 2017 as classical computing reached its limits in optimizing tasks such as energy demand forecasting and electric vehicle charging. The EDF Quantum Project Team focuses on solving complex energy sector problems through quantum technologies.

• EDF partnered with Pasqal to apply quantum computing in optimization challenges, starting with smart charging for electric vehicles. This collaboration has since expanded to simulate physical phenomena affecting energy production, such as modeling the environmental conditions impacting wind farms and photovoltaic plants, and simulating material degradation in nuclear plants.

• EDF plans to continue developing its quantum capabilities, focusing on testing quantum algorithms on real quantum machines and scaling up quantum applications to address broader operational challenges. The goal is to determine the extent of quantum advantage and integrate it into core energy management practices.

🤖 Quantum Neural Network Using Hamming-Weight Circuits Improves Upon Image Classification, Achieving Faster Processing and Fewer Parameters

QUICK BYTE: Scientists from Sorbonne Universite, Naval Group in France, and QC Ware introduce a subspace-preserving quantum convolutional neural network architecture that uses Hamming-weight preserving quantum circuits for efficient image classification tasks, demonstrating polynomial speed-ups and reduced parameters compared to classical CNNs.

DETAILS

• The proposed QCNN architecture, based on Hamming-weight preserving quantum circuits, shows competitive performance with classical convolutional neural networks, achieving comparable classification accuracy with fewer parameters on datasets such as MNIST, Fashion MNIST, and CIFAR-10.

• The QCNN uses Hamming-weight preserving quantum convolutional layers, pooling layers, and dense layers, reducing computational complexity while maintaining structure and avoiding the barren plateau phenomenon common in quantum neural networks. The quantum convolution layers use reconfigurable beam splitter gates for encoding, while the pooling layers introduce non-linearities via measurement operations.

• This quantum architecture demonstrated polynomial speed-ups over classical models, especially in high-dimensional tasks such as image classification, by reducing the number of parameters and eliminating the need for vectorization commonly required in classical models. It illustrates quantum computing's potential in deep learning applications with near-term quantum devices.

• Simulations show the quantum architecture outperforming classical CNNs on the MNIST dataset, achieving a training accuracy of 93.79% and testing accuracy of 86.79% with 755 parameters, compared to 91.33% training and 84.59% testing accuracy with 990 parameters for the classical CNN.

Researchers from Lawrence Berkeley National Laboratory and the University of Massachusetts Amherst have introduced QubiCSV, an open-source platform that streamlines the management, visualization, and version control of quantum calibration and characterization data. Designed to address challenges in handling large volumes of quantum data and promoting collaboration across teams, QubiCSV includes features like real-time data management, data versioning, and advanced visualization tools to track qubit control parameters such as drive frequencies and gate operations.

Quantum Computing Inc. has extended its Cooperative Research and Development Agreement (CRADA) with Los Alamos National Laboratory to further develop its Dirac-3 quantum optimization machine. Dirac-3, which uses nonlinear quantum optics, is designed to solve highly complex optimization problems beyond traditional QUBO problems, addressing multi-body interactions and high-dimensional discrete variables. QCi's Dirac-3 is already being applied to challenges in energy grid management, offshore wind optimization, and signal processing, and the company is collaborating with NASA and the U.S. Department of Defense on additional projects.

Archer Materials has developed a scalable carbon film with electron spin lifetimes exceeding 400 nanoseconds, which offers advantages for quantum device manufacturing. Using a proprietary chemical vapor deposition process, the films can be processed with standard semiconductor techniques, overcoming previous challenges with carbon nano-onions. This development accelerates the production of quantum devices, and collaboration with EPFL aims to further refine spin lifetime control in these films. Archer expects this to contribute to faster observation of quantum phenomena and rapid quantum device development.

Scientists from the Federal University of Rio de Janeiro and the Technology Innovation Institute in Abu Dhabi introduce a hybrid quantum-classical framework for simulating matrix functions, improving early fault-tolerant quantum hardware usage. The method utilizes randomization over Chebyshev polynomial approximations and a modified Hadamard test to enhance circuit efficiency and reduce sensitivity to noise. Applied to tasks like partition function estimation, linear system solvers, and ground-state energy estimation, the technique offers runtime reductions and noise resilience.

Arqit Quantum Inc. has entered into a securities purchase agreement with existing shareholders for the sale of 5.44 million ordinary shares at $2.50 per share, generating approximately $13.6 million in gross proceeds. Concurrently, the company will issue unregistered warrants to purchase an equivalent number of shares, exercisable only under specific conditions, including a $5.00 share price threshold sustained for 60 consecutive trading days. The proceeds will be used for general corporate purposes, and the offering is expected to close by October 9, 2024.

LISTEN

On the most recent episode of The New Quantum Era, Sebastian Hassinger and Kevin Rowney sit down with Martin Schultz, Professor at TU Munich and board member of the Leibniz Supercomputing Center. They discuss the integration of quantum and classical computing to build practical quantum solutions, focusing on the Munich Quantum Valley initiative. Key topics include building a hybrid quantum-classical infrastructure, efficient resource scheduling, and the importance of open-source collaboration for future quantum developments.

ENJOY

Whether it brings you enjoy, or you’d rather use a literal potato, Barcelona-based PhD student Luke Mortimer has thoroughly entertained the quantum community today. He ported the classic game DOOM to quantum computers, naming the project "Quandoom." However, no current quantum computer can handle the steep requirements of 72,376 qubits and 80 million gates, so the game runs on a QASM simulator for classical computers. Quandoom can achieve 10-20 FPS on a modest laptop, though it lacks color, sound, and features beyond the first level. Lumorti's project involves over 8,000 lines of C++ code and uses quantum registers, with potential for the source code release if enough interest is shown.

WATCH

A lesson on the concept of purification in quantum information and fidelity, including Schmidt decompositions, unitary equivalence of purifications, and Uhlmann's theorem: