Collins Conference Room
Working Group
  US Mountain Time
 

Our campus is closed to the public for this event.


Abstract. Spin glasses are the archetypal native benchmark problem for any novel algorithm or computing architecture. The fact that a spin-glass-like Hamiltonian on a non-planar topology falls typically under the NP-hard complexity class is the main reason why these systems are of interest to the algorithm development and benchmarking community. In recent years, highly tunable spin-glass benchmarks have been developed to probe the efficiency of both classical and novel quantum architectures, like the D-Wave Inc. DW2X device. Because the number of qubits in quantum annealers is doubling every two years, soon it will be difficult to generate validation libraries of spin-glass problems that are also difficult to solve while being good benchmarks. As such, planting solutions – the process in which one first generates the solution to a problem and then builds the corresponding cost function around it – has gained much interest recently. Approaches based on the planting of random-constraint-satisfaction-problem formulas are straightforward, however the produced benchmark problems are too easy to challenge state-of-the-art algorithms and computing architectures. The generation of planted solutions is intimately related to the inverse Ising problem where, again, a model Hamiltonian is built around an existing data set. In this working group we attempt to tackle these difficult problems using novel approaches.

Purpose: 
Research Collaboration
SFI Host: 
Helmut Katzgraber

More SFI Events