Solving Big Math Problems with Small Quantum Computers

This patent describes how a classical computer can break down large mathematical problems into smaller pieces that even limited quantum computers can solve, then combine the results.

What it covers

This patent describes a method for solving complex mathematical problems, called 'objective functions,' that are too large for current quantum computers. First, a classical computer obtains an objective function that has more variables than the quantum computer has 'logical qubits' (Claim 1). Next, the classical computer uses machine learning to break down this big problem into several smaller 'sub-problems,' each small enough for the quantum computer to handle (Abstract, Claim 1). The quantum computer then solves each sub-problem by initializing its qubits, applying changes (perturbations), and measuring the results multiple times ('shots') (Claim 1). From these measurements, an 'expectation value' is determined for each sub-problem, leading to its solution (Claim 1). Finally, the classical computer gathers all the sub-problem solutions to find the overall solution to the original big problem and stores it (Claim 1). For example, if you have a complex optimization problem with 100 variables, but your quantum computer only has 10 logical qubits, this method would break the 100-variable problem into several 10-variable sub-problems.

What it doesn't cover

• —Does not cover solving mathematical problems that involve fewer variables than the quantum computer has logical qubits, meaning the problem fits directly on the quantum computer (Claim 1).
• —Does not cover methods that solve the entire objective function directly on a quantum computer without first decomposing it into smaller sub-problems (Claim 1).
• —Does not cover purely classical computer systems solving large problems without involving a quantum computing system for sub-problem solutions (Claim 1).
• —Does not cover determining sub-problem solutions without using an 'expectation value' derived from a set of raw outputs from multiple 'shots' (Claim 1).
• —Does not cover decomposition methods that do not involve machine learning, as described in the abstract.

The clever bit

The core innovation is intelligently breaking down a problem too big for a quantum computer into smaller, manageable pieces using machine learning, then reassembling the quantum-derived solutions. This overcomes the major hurdle of limited qubit availability in current quantum hardware.

Why it matters

Quantum computers today have a limited number of stable qubits, which restricts the size and complexity of problems they can solve. This patent addresses a fundamental challenge in quantum computing by providing a way to tackle larger, more commercially relevant problems despite these hardware limitations. It enables a hybrid approach, combining the strengths of classical computers for problem management with quantum computers for specific computational tasks. This approach is crucial for expanding the practical applications of quantum computing in the near term.

Real-world examples

• 1.Optimization problems in logistics or finance
• 2.Drug discovery simulations
• 3.Materials science research
• 4.Quantum machine learning applications
• 5.Hybrid quantum-classical algorithms