Dataset References¶
Below contain the references to the datasets we gathered on this website.
[1] Variational Neural Annealing
VNA proposes a parameterized annealing model to stochastically search for ground states of the Ising model. It features fully connected spin glass instances, as well as Edwards-Anderson (EA) and Wishart Planted Ensemble (WPE) graph instances of the Ising model.
[2] Searching for Spin Glass Ground States Through Deep Reinforcement Learning
This paper introduces DIRAC, a deep reinforcement learning framework that can be trained on small-scale spin glass instances and applied to arbitrarily large ones. DIRAC has found success in scalability compared to other methods and has been tested on 2D, 3D, and 4D EA spin-glass instances.
[3] Finding the Ground State of Spin Hamiltonians with Reinforcement Learning
This paper demonstrates that an RL agent is able to surpass the performance of standard heuristic temperature schedule for two classes of Hamiltonians. They show the performance of their implementation by training on weak-strong clusters (bipartite graph with two fully connected clusters) and nearest-neighbor square spin glasses.
[4] Quantum-enhanced Markov Chain Monte Carlo
This paper provides an algorithm that leverages MCMC to sample from the Boltzmann distribution of classical Ising models. It performs testing analysis on spin-glass Ising models ranging from 3 to 10 spins, and each set of spins featuring both a fully connected and line connected configuration.
[5] Wavelength-Division Multiplexing Optical Ising Simulator
This paper uses a spatial photonic Ising machine to compute Ising Hamiltonians of programmable Ising models. To demonstrate the programming capabilities, testing is performed on $pm$ J models, Sherrington-Kirkpatrick (SK) models, and locally connected $J_1-J_2$ models
[8] Beyond-classical computation in quantum simulation
This paper uses seeks to establish quantum computing’s capability of handling complex computational problems, within the field of quantum simulation. With the problem scope focusing on the Ising model, they compare their QPU’s results against classical algorithms (Schrodinger equation dependent) on various spin-glass models. Specifically, testing was performed on 2D cylindrical spin-glass, 3D cubic spin-glass, 3D diamond spin-glass, and bi-clique spin-glass instances.
[9] Quantum-enhanced greedy combinatorial optimization solver
This paper introduces a quantum heuristic optimization algorithm for combinatorial optimization problems. Attempting to overcome noise constraints by reducing the problem to a classical greedy problem. Specifically, this heuristic algorithm is tested on Sherrington-Kirkpatrick spin-glass problems.
[11] Biq Mac Library
A dataset for QUBO and Max-Cut instances curated for the development of a QUBO and Max-Cut solver (Biq Mac). Within this dataset contains Ising model instances, specifically, Toroidal grid graphs and Ising chain instances.
[12] On Computational Complexity of 3D Ising Spin Glass: Lessons from D-Wave Annealer
This paper searches for the ground state of 3D spin-glass instances using annealing methods on the D-Wave quantum annealer. To demostrate the NP-hard complexity of 3D spin-glass instances, they performed testing on 3D instances ranging from 3558 to 5627 spins.
[13] A high-bias, low-variance introduction to Machine Learning for physicists
[14] Solving Statistical Mechanics Using Variational Autoregressive Networks