Quantum Sampler Training via Classically Simulable Circuits
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current quantum sampler training methods are resource-intensive and time-consuming due to the slow clock-rate of quantum devices, with 99% of computational resources spent on training and only 1% on sample generation, making it inefficient to train quantum models that require exponential classical computing resources.
Innovation Solution
Classically train quantum models using quantum circuits that can be simulated efficiently on classical computers, while the generation of samples remains classically hard, utilizing circuits like IQP circuits that allow polynomial-time probability estimation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If quantum circuits are executed on hardware quantum registers for training, then quantum advantage can be achieved for sample generation, but training time and resource consumption increase significantly due to slow clock-rate
Solution Approach 1:
The training process is segmented into two distinct phases: a classical training phase where probability amplitudes are computed and optimized using classical computers, and a quantum sampling phase where the optimized circuit is executed on quantum hardware. This segmentation allows the computationally intensive optimization to be performed efficiently classically, while only the final sampling utilizes quantum resources, thereby reducing overall training time while maintaining quantum advantage.
Solution Approach 2:
A classical computer serves as an intermediary between the quantum circuit definition and the quantum hardware execution. The classical computer computes probability amplitudes, optimizes circuit parameters, and prepares the final circuit for quantum execution. This intermediary role allows efficient classical optimization without requiring continuous quantum hardware involvement, significantly reducing training time.
2Manufacturing precision
If parameterized quantum circuits are used for training with gradient optimization, then the target probability distribution can be achieved, but computational resources scale exponentially with the number of qubits
Solution Approach 1:
The patent uses classical computers to create a computational copy or simulation of the quantum circuit's probability amplitude calculations. By computing these amplitudes classically during the training phase, the system avoids the need for exponential quantum resources while still achieving accurate optimization of the target probability distribution. The optimized circuit can then be efficiently executed on quantum hardware for sampling.
3Measurement precision
If multiple samples are generated for gradient estimation, then training accuracy improves, but the slow clock-rate of quantum devices makes the process very resource intensive
Solution Approach 1:
The patent replaces the quantum mechanical sampling process with a classical computational approach for the training phase. Instead of using quantum hardware to generate multiple samples for gradient estimation, the system uses classical computers to compute probability amplitudes and estimate gradients efficiently. This substitution eliminates the bottleneck imposed by quantum device clock-rates while maintaining the accuracy needed for reliable training.
Data Source
AI summary
A method for training and sampling a quantum model comprises: training a quantum model as a quantum sampler configured to produce samples which are associated with a predetermined target probability distribution and which are exponentially hard to compute classically, the training including classically computing probability amplitudes associated with an execution of a first parameterized quantum circuit that defines a sequence of gate operations for a quantum register and optimizing parameter(s) of the first parameterized quantum circuit based on the classically computed probability amplitudes; and, executing a sampling process using the hardware quantum register including determining an optimized quantum circuit based on the optimized parameter(s) and the first parameterized quantum circuit or a second parameterized quantum circuit, which is related to the first parameterized quantum circuit, and executing the optimized quantum circuit on the hardware quantum register and generating a sample by measuring the output of the hardware quantum register.