Abstract:Probabilistic Bit (P-Bit) device serves as the core hardware for implementing Ising computation. However, the severe intrinsic variations of stochastic P-Bit devices hinder the large-scale expansion of the P-Bit array, significantly limiting the practical usage of Ising computation. In this work, a behavioral model which attributes P-Bit variations to two parameters {\alpha} and {\Delta}V is proposed. Then the weight compensation method is introduced, which can mitigate {\alpha} and {\Delta}V of P-Bits device variations by rederiving the weight matrix, enabling them to compute as ideal identical PBits without the need for weights retraining. Accurately extracting the {\alpha} and {\Delta}V simultaneously from a large P-Bit array which is prerequisite for the weight compensation method is a crucial and challenging task. To solve this obstacle, we present the novel automatic variation extraction algorithm which can extract device variations of each P-Bit in a large array based on Boltzmann machine learning. In order for the accurate extraction of variations from an extendable P-Bit array, an Ising Hamiltonian based on 3D ferromagnetic model is constructed, achieving precise and scalable array variation extraction. The proposed Automatic Extraction and Compensation algorithm is utilized to solve both 16-city traveling salesman problem(TSP) and 21-bit integer factorization on a large P-Bit array with variation, demonstrating its accuracy, transferability, and scalability.

What problem does this paper attempt to address?

### What problem does this paper attempt to solve? This paper aims to solve the problem of inaccurate Ising calculations in large - scale P - Bit (probability - bit) arrays due to the inherent variations of P - Bit devices. Specifically: 1. **Inherent variations of P - Bit devices**: - P - Bit devices are the core hardware for implementing Ising calculations, but their inherent randomness leads to significant differences between devices. - These differences make it difficult to expand large - scale P - Bit arrays, thus limiting the practical applications of Ising calculations. 2. **Weight compensation method**: - The paper proposes a behavioral model, attributing the P - Bit variations to two parameters, \(\alpha\) and \(\Delta V\). - A weight compensation method is introduced. By re - deriving the weight matrix, the impact of these variations is mitigated, enabling P - Bit devices to calculate like ideal P - Bit devices without re - training the weights. 3. **Automatic extraction algorithm**: - Accurately extracting \(\alpha\) and \(\Delta V\) simultaneously from large - scale P - Bit arrays is a prerequisite for implementing weight compensation, which is a crucial and challenging task. - To this end, the paper proposes a new automatic variation extraction algorithm based on Boltzmann machine learning, which can extract the variation of each P - Bit from large - scale P - Bit arrays. 4. **Application of 3D ferromagnetic model**: - To achieve accurate and scalable array variation extraction, the paper constructs a 3D ferromagnetic model based on the Ising Hamiltonian. - This model ensures the uniformity of the density matrix and the sparsity of the system, thereby improving the accuracy and scalability of variation extraction. 5. **Verification and application**: - The paper demonstrates the effectiveness, transferability, and scalability of the proposed automatic extraction and compensation algorithms in solving the 16 - city traveling salesman problem (TSP) and the 21 - bit integer factorization problem. In summary, by introducing new behavioral models, weight compensation methods, and automatic variation extraction algorithms, this paper solves the impact of P - Bit device variations on the accuracy of Ising calculations and provides an effective solution for the practical application of large - scale P - Bit arrays.

Automatic Extraction and Compensation of P-Bit Device Variations in Large Array Utilizing Boltzmann Machine Training

Training Deep Boltzmann Networks with Sparse Ising Machines

Mean-Field Assisted Deep Boltzmann Learning with Probabilistic Computers

Time-Division Multiplexing Ising Computer Using Single Stochastic Magnetic Tunneling Junction

Quantitative Evaluation of Hardware Binary Stochastic Neurons

All-to-all reconfigurability with sparse and higher-order Ising machines

Probabilistic computing with NbOx metal-insulator transition-based self-oscillatory pbit

Lateral approach to the os calcis.

Energy-efficient superparamagnetic Ising machine and its application to traveling salesman problems

Improved Sparse Ising Optimization

Scalable In-Memory Clustered Annealer With Temporal Noise of Charge Trap Transistor for Large Scale Travelling Salesman Problems

Solving Boltzmann Optimization Problems with Deep Learning

A Variation-Tolerant Continuous-Time Ising Machine with Edram-Based Spin Interaction and Leaked Negative Feedback Annealing

p-Bits for Probabilistic Spin Logic

Stochastic p-Bits Based on Spin-Orbit Torque Magnetic Tunnel Junctions

30.5 A Variation-Tolerant In-eDRAM Continuous-Time Ising Machine Featuring 15-Level Coefficients and Leaked Negative-Feedback Annealing

Machine Learning Quantum Systems with Magnetic p-bits

Toward Practical Benchmarks of Ising Machines: A Case Study on the Quadratic Knapsack Problem

Variation-Resilient FeFET-Based In-Memory Computing Leveraging Probabilistic Deep Learning

Probabilistic Modeling for Continuous EDA with Boltzmann Selection and Kullback-Leibeler Divergence