Quantum Computing
Fujitsu believes that the organizations leading the quantum race will be those that have invested significant time in researching and developing quantum solutions to address intractable challenges within their domains. To support this vision, Fujitsu offers a comprehensive suite of technologies and solutions that integrate its world-class superconducting quantum computer, one of the world’s largest quantum simulators, and proprietary AI technologies—all accessible through a unified platform.
Fujitsu is committed to advancing quantum computing, with the goal of building a large-scale superconducting quantum computer featuring over 10,000 physical qubits by 2030, and achieving 1,000 logical qubits by 2035[1].
Quantum Machine Learning
- Quantum Machine Learning (QML) merges quantum computing principles with machine learning techniques to solve problems that may be intractable using classical methods alone. This includes:
- Enhancing existing machine learning models with quantum speedup
- Designing hybrid algorithms that combine classical and quantum computation
- Leveraging quantum-native approaches for analyzing quantum data
- FRIPL’s research in QML is focused on developing algorithms that can run on Fujitsu’s quantum simulators and hardware, with long-term potential for applications in optimization, language processing, and high-dimensional data analysis.
Quantum Algorithm
- FRIPL focuses on creating quantum algorithms that are not only theoretically powerful but also practically applicable using Fujitsu’s quantum hardware and simulators. Areas of focus include:
- Variational Quantum Algorithms (VQAs) for use in optimization, quantum chemistry, and simulation
- Fault-Tolerant Quantum Computation (FTQC) algorithms for complex problems like database search and prime factorization
- While many quantum algorithms promise exponential speedups, most require a large number of qubits and high circuit depth. FRIPL is actively exploring hybrid quantum-classical algorithms that can function within the limitations of current Noisy Intermediate-Scale Quantum (NISQ) devices to deliver near-term advantage.
Quantum Error Correction
Quantum Error Correction (QEC) is essential to ensure the reliability of quantum systems by addressing errors caused by quantum noise. Since quantum information cannot be copied directly due to the no-cloning theorem, QEC relies on distributing quantum states across multiple qubits and using syndrome measurements for error detection and correction.
FRIPL is researching the design and application of robust QEC codes that are critical for the transition to large-scale fault-tolerant quantum systems.
Publications
- Model selection in hybrid quantum neural networks with applications to quantum transformer architectures
- Introduced the Quantum Bias-Expressivity Toolbox (QBET) for efficient pre-screening of hybrid quantum classical models without training using lean metrics for Simplcity Bias (SB) and Expressivity (EXP)
- The toolbox was applied to a variety of real-world tasks like molecular graph generation and image classification using quantum transformers, to show strong correlation between SB and EXP, and downstream performance metrics. Importantly, using QBET we identify several hybrid models showed which show improved performance over classical counterparts.
- Introduced the Quantum Bias-Expressivity Toolbox (QBET) for efficient pre-screening of hybrid quantum classical models without training using lean metrics for Simplcity Bias (SB) and Expressivity (EXP)
- Quantum Generative Adversarial Autoencoders: Learning latent representations for quantum data generation
- Introduced the Quantum Generative Adversarial Autoencoder (QGAA), a quantum model for generation of quantum data, consisting of (a) an Quantum Autoencoder (QAE) to compress quantum states, and (b) Quantum Generative Adversarial Network (QGAN) to learn the latent space of the trained QAE.
- The model has been demonstrated using two representative examples: generation pure entangled states and parametrized molecular Hamiltonians for Hydrogen and Lithium Hydride.
- Enhancing variational quantum algorithms by balancing training on classical and quantum hardware
- Introduced new quantum circuit ansatz and training methods for distributing the gradient-estimation workload between quantum and classical hardware
- Reduced quantum hardware usage by up to 60% and error by an order of magnitude for ground-state estimation, with 2.8% increase in accuracy for classification task
- Check out our technical blog here!
- LinkedIn post to know more!
- Introduced new quantum circuit ansatz and training methods for distributing the gradient-estimation workload between quantum and classical hardware
- Meta-learning of Gibbs states for many-body Hamiltonians with applications to Quantum Boltzmann Machines
- Introduced a collective optimization technique for Gibbs state preparation, with better accuracy and resource requirements compared to existing variational methods
- Application of above method to Quantum Boltzmann Machines improves existing workflow with up to 30x reduction in time
- Introduced a collective optimization technique for Gibbs state preparation, with better accuracy and resource requirements compared to existing variational methods
Patents
Loading component...
Conferences
FRIPL is actively engaging with the quantum community all over the world through conferences, workshop and collaborations.
