Accenture Labs, the IT services firm’s R&D arm, is continuing its push into quantum computing, with a focus on identifying appropriate applications for this emerging technology.
Patent activity over the past few months sheds light on the organization’s direction. In December 2018, Accenture announced it received a U.S. patent for its method of harnessing quantum computing and classical computing to address business challenges. The patent covers a “multistate quantum optimization engine” that runs multiple simulations at the same time to find the best way to solve a particular problem.
In July 2019, Accenture said it was awarded a patent for a “quantum computing machine learning module.” The module trains AI models to determine which computational tasks should be handled via quantum computing, as opposed to classical computing methods. The module then routes a computational task to the appropriate computing resource.
Accenture’s latest patented quantum computing approach aims to find the cost-performance sweet spot for a particular workload.
“The patent is really about understanding and creating systems that allow us to deal with the computational diversity that is available today,” said Carl Dukatz, senior manager of Accenture Labs systems and platforms R&D. “Specifically, this one doubles down on adding quantum computing to the mix.”
Dukatz is among the inventors behind the patent.
Quantum computing offers the potential for enormous gains in processing power. The ability for quantum information, or qubits, to exist simultaneously in multiple states, a property called superposition, is central to the jump in performance.
IBM, which unveiled Q System One earlier this year, and D-Wave Systems are among the vendors developing quantum hardware. Market researchers, however, anticipate the advent of mainstream applications running on quantum computers to be at least three years away.
In the meantime, quantum-inspired computers are appearing as a kind of evolutionary step between classical and quantum computers. Such computers apply quantum concepts, but use traditional computing architectures that don’t require the supercooling of a quantum computer or involve other limitations, such as instability.
Dukatz said he envisions a “slow transition” from using strictly classical machines to the introduction of quantum-inspired systems and then moving on to the eventual replacement of some classical systems by quantum computers.
Accenture’s objective is “to monitor and move through the transitional period as seamlessly as possible,” he said, noting the quantum computing machine learning module is one tool for doing that.
How it works
Accenture’s module uses “machine learning techniques to determine when and how to leverage the power of quantum computing,” according to the company’s patent. The machine learning module receives computational tasks and routes them to external computing devices. Those devices could include different types of quantum computers, such as quantum annealers, quantum simulators and universal gate computers, along with one or more classical computers.
Carl DukatzSenior manager of Accenture Labs systems and platforms R&D
A machine learning model within Accenture’s module is trained to determine how to best direct a given workload. That learning process is based on training data collected from previous computational tasks. Training data can include information on the specific properties of the devices available to crunch workloads and information on the size and complexity of the computational job. The training data may also take into account the time it takes a computing device to come up with a solution and the computational cost of doing so.
The machine learning module itself may be based on a classical or quantum computer, as the latter become more readily available. Dukatz said an implementation today would use an artificial neural network running on a classical machine, while a future implementation could use a support vector machine running on quantum hardware. Artificial neural networks and support vector machines are algorithms that can be used in machine learning.
The machine learning system may pick out which quantum platform is best suited for a job, or find that none of them are.
“(A) computational task may be too complex to implement using quantum devices, or may require a large classical overhead, meaning that using the quantum device to solve the computational task is either slower or more costly than using a classical device to solve the computational task,” Accenture’s patent noted.
A focus on quantum computing applications
Accenture’s overarching focus is on identifying appropriate applications for the emerging technology. Biotechnology is one such area. Accenture Labs and quantum software firm 1QBit have been working with biotech company Biogen to explore how quantum techniques can be used to compare and match molecular structures.
“Right now, there’s a lot of focus on chemistry,” Dukatz said. Examples like Biogen show quantum-inspired technology provides “better matching between molecular structures than some of the traditional methods,” he said. “There’s a space for quantum-inspired machines until quantum is up to scale to run something like that.”
But quantum computing applications such as optimizing a financial portfolio or a supply chain system could hold even bigger potential on the business side.
“The biggest opportunity is in optimization,” said Teresa Tung, managing director in Accenture Labs, responsible for its global systems and platforms R&D group.
Accenture’s machine learning module “is really about finding the right opportunity to apply quantum in business applications, not just scientific applications,” she added.
This is a syndicated post. Read the original post at Source link .