Downsizing Machine Learning Models by Optimization through Ising Models

This paper introduces a technique to enhance the efficiency of quadratic machine learning models, particularly Field-Aware Factorization Machines (FFMs) handling binary data. Our approach strategically reduces model size through optimized feature selection based on the Ising model, maintaining comparable accuracy to the original model. By exploiting the adjustability of FFM's cross-term weights during a novel two-stage training process, we demonstrate a significant improvement in overall computational efficiency.