Hardware-efficient AI and ML Current State-of-the-Art:
Combinatorial optimization problems (COPs) involve optimizing systems with discrete variables, posing significant challenges due to the inapplicability of traditional gradient-based algorithms. Moreover, these problems are considered to be NP-hard. Recently, Ising machines—specialized computational systems inspired by the Ising model in physics—have emerged as promising tools for solving COPs. These machines map problems onto a network of spins, allowing them to leverage physical processes or specialized hardware to explore solutions efficiently.
Research Gap:
Despite their potential, Ising machines face two critical limitations. First, they often disregard hardware inaccuracies, such as noise and variability in physical implementations, which can degrade solution quality. Second, their scalability remains limited, restricting their application to relatively small problems.
Research Goal:
This research aims to design and implement an Ising machine capable of solving large-scale combinatorial optimization problems with high accuracy. By focusing on overcoming hardware inaccuracies and scaling limitations, the study seeks to push the boundaries of Ising machines' capabilities, enabling them to address complex, real-world COPs at large scales.
