Solving complex combinatorial optimization problems faster with improved ising machine technology
"more scalable... halve required spin–spin interactions using matrix-folding... large-scale combinatorial optimization... smaller circuit footprint... half of interaction matrix redundant, can be omitted... divided matrix into 4 sections, halving each, alternatively preserving either top or bottom halves of each submatrix. then folding/ rearranging remaining elements into rectangular shape... 384 spins on 8 FPGA chips... performance ratio 400X better... data center/ power grid efficiency, faster molecular simulations"
Comments
Post a Comment