QC Ware defines itself as a quantum computing-as-a-service company that builds enterprise solutions to run on quantum computing hardware. It recently announced several significant breakthroughs in quantum machine learning (QML).
What is machine learning?
Machine learning is a subset of artificial intelligence. It has evolved from simple work in the 1950s to today’s deep learning that uses sophisticated training and neural networks.
Machine learning allows a computer to make decisions based on what it learns about the characteristics of large data sets. We use it to classify and find similarities in data or images or text, as well as cluster grouping for tasks such as forecasting, prediction and anomaly detection. As a machine learning algorithm continues to learn from data, it gains more and more insights about the data, and, as a consequence, its predictions become more and more accurate.
Quantum machine learning
Quantum machine learning is at the intersection of classical machine learning and quantum computing. However, because of current limitations in quantum computing technology, useful machine learning is primarily confined to the realm of classical computing.
There are several reasons that quantum machine learning is not currently possible:
- There is no working quantum random access memory (QRAM) available. QRAM is the quantum equivalent of classical RAM, and it is necessary to convert large amounts of classical data into corresponding quantum states. Even if QRAM were available today, the number of qubits required for QRAM would far exceed the number available in any current quantum computer.
- And two, putting QRAM qubit requirements aside, today’s noisy quantum computers have an insufficient number of qubits to perform the complex computations needed for machine learning.
Why QC Ware’s new developments are significant
QC Ware researchers have developed Data Loaders, a QRAM alternative that can efficiently and easily load classical data onto quantum hardware.
They are also an efficient method to perform distance estimations on a quantum computer. Distance estimation is an algorithm used in machine learning that tries to group each data point with other points or clusters with similar properties.
QC Ware’s Data Loaders are available on its cloud platform called Forge.
Forge provides access to quantum algorithms on quantum hardware, simulators on quantum hardware, and classical simulators that allow enterprises to build, edit and implement quantum algorithms.
“QC Ware estimates that with Forge Data Loaders, the industry’s 10-to-15-year timeline for practical applications of QML will be reduced significantly,” said Yianni Gamvros, Head of Product and Business Development at QC Ware. “What our algorithms team has achieved for the quantum computing industry is equivalent to a quantum hardware manufacturer introducing a chip that is 10 to 100 times faster than their previous offering. This exciting development will require business analysts to update their quad charts and innovation scouts to adjust their technology timelines.”
QC Ware has developed two types of Data loaders, a parallel Data Loader and an optimized Data Loader, both of which convert classical data into quantum states for machine learning applications. And as mentioned, an optimized Distance Estimation algorithm is also available.
The table below, furnished by QC Ware, shows a theoretical comparison of the differences for loading data points with a thousand features each.
There is a significant difference between the required number of qubits and circuit depth for QRAM hardware and QRAM quantum-inspired circuits compared to the QC Ware Data Loader options.
“To gain performance speedups on near-term quantum computers, it’s important to keep pushing the boundaries of what is possible with current hardware and current algorithms,” said Iordanis Kerenidis, Head of Quantum Algorithms International at QC Ware. “We are constantly striving to make fewer qubits and shallower circuits do more through innovative algorithms.”
GPU acceleration for simulation plus new quantum algorithms
In addition to the machine learning Data Loaders, the latest release of Forge also includes tools for GPU acceleration of quantum algorithms. With GPU acceleration, depending on the number of shots, algorithm testing can be reduced to seconds instead of hours.
The new Forge release includes various turnkey algorithm implementations. Considering the number of new quantum developers expected to enter the field, and the ease of using algorithms, this is a big plus. The D-Wave charts I’ve seen are particularly impressive. Each implementation offers unique performance advantages and capabilities:
- Users can now run quantum classification, regression, and clustering algorithms on larger problems than what was previously possible as the implementations use the Forge Data Loaders and Distance Estimation. (Forge contains classification and clustering examples that can run on a simulator with user-specified data sets)
- Improved quantum annealing performance on D-Wave, by 10 to 100 times for larger problems
- Optimization of algorithm parameters for optimization algorithms
Also, the algorithms can be executed on any of the following backends:
- Classical CPU simulators
- Classical GPU simulators (NVIDIA GPUs provisioned on AWS)
- Hardware on Amazon Braket, which includes access to IonQ and Rigetti hardware
- D-Wave Systems hardware
Because useful quantum machine learning is still many years away, QC Ware’s new products are definitely a positive contribution to not only QML, but quantum computing in general.
The new Forge additions will promote more research which will ultimately push the development of QML forward and at a faster pace.