Description
Simulating and finding the ground state of Ising spin networks with physical systems is a promising approach with which to solve mathematically intractable problems such as combinatorial optimization or neural network training. Two-dimensional Ising spin networks in particular are paradigmatic with respect to many statistical systems and serve as a common benchmark. We use an artificial spin-network of 1936 coherently coupled degenerate optical parametric oscillators to realize and solve large-scale 2D Ising models in a coherent Ising machine with an opto-electronic feedback system. Due to the fast optical dynamics, the correct ground state can be found within just a few milliseconds. Comparisons with Monte Carlo simulations reveal that coherent Ising machines operating at room temperature implement low temperature spin systems, which makes them inherently suitable for optimization tasks. Beyond that, we also propose a way to implement a temperature control, which allows for fast sampling of Boltzmann distributon functions of the 2D Ising model at arbitrary temperatures. We show how this can be applied to the training of neural network.| Period | 18 Jun 2018 → 20 Jun 2018 |
|---|---|
| Event title | Nonlinear Dynamics in Semiconductor Lasers |
| Event type | Workshop |
| Location | Berlin, Germany |
| Degree of Recognition | International |