Dynamic Range Reduction via Branch-and-Bound

Thore Gerlach, Nico Piatkowski

TL;DR: Investiagtion of a fully principled precision reduction oriented Branch-and-Bound algorithm for hardware accelerators such as FPGAs, GPUs and Quantum Computers.

Abstract:

A key strategy to enhance specialized hardware accelerators, such as GPUs and FPGA, is to reduce the numerical precision in arithmetic operations, which increases processing speed and lowers latency---both crucial for real-time AI applications.
In this work, we consider NP-hard Quadratic Unconstrained Binary Optimization (QUBO) problems, which arise in machine learning and beyond.
We show that these problems often require high numerical precision, posing challenges for hardware solvers.
We introduce a principled Branch-and-Bound algorithm for reducing the precision requirements of QUBO problems by utilizing the dynamic range as a measure of complexity. Experiments demonstrate that our method reduces the dynamic range in subset sum, clustering, and vector quantization problems, thereby increasing their solvability on actual quantum annealers and FPGA-based digital annealers.

Primary Area: Optimization->Discrete and Combinatorial Optimization

Keywords: QUBO, Precision Reduction, Dynamic Range, Quantum Annealing, Hardware Accelerators

Submission Number: 3295