Sampling Server Offloading for Quantum Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning systems face limitations in processing problems with size and/or connectivity greater than what analog processors can handle, as they often require more computation devices and couplers than available in these processors.
Innovation Solution
A computational system that includes a digital processor core and a sampling server, which receives initial parameters for a machine learning process, generates samples, and provides these samples for further iterations, allowing concurrent execution with the machine learning process. This system utilizes a quantum processor to draw samples from a Boltzmann distribution and performs post-processing on the samples before sending them back for use in the machine learning algorithm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If sampling operations are performed within the same analog processor as the machine learning algorithm, then the system can process problems within the processor's capacity, but the running time of the machine learning algorithm increases due to sequential execution
Solution Approach 1:
The system divides the computation into two independent segments: the machine learning algorithm executes on the analog processor while sampling operations execute on a separate digital processor. This segmentation allows both operations to proceed simultaneously without interfering with each other, resolving the time loss issue while maintaining processing speed.
Solution Approach 2:
A sampling server acts as an intermediary component that communicates between the analog processor and the digital processor. The server receives parameters from the analog processor, manages the sampling operations on the digital processor, and returns samples to complete the machine learning algorithm, enabling efficient coordination between the two processors.
2Adaptability or versatility
If more computation devices and couplers are used to handle larger problem graphs, then the system can process more complex problems, but the hardware resource requirements increase
Solution Approach 1:
The digital processor serves multiple functions: it performs sampling operations, processes samples, and communicates with the analog processor. This multi-functionality reduces the need for dedicated hardware components for each function, thereby reducing overall device complexity while maintaining the ability to handle complex problem graphs.
Solution Approach 2:
Instead of physically expanding the analog processor to handle larger problems, the system creates a digital copy that performs sampling operations. This virtual copying approach allows the system to handle more complex problems without increasing the physical hardware complexity of the analog processor itself.
Data Source
AI summary
A digital processor runs a machine learning algorithm in parallel with a sampling server. The sampling sever may continuously or intermittently draw samples for the machine learning algorithm during execution of the machine learning algorithm, for example on a given problem. The sampling server may run in parallel (e.g., concurrently, overlapping, simultaneously) with a quantum processor to draw samples from the quantum processor.