Ising vs. QUBO encodings impact Boltzmann machine learning convergence

A new research paper evaluates the performance of Ising and QUBO variable encodings in Boltzmann machine learning. The study found that QUBO encodings can lead to ill-conditioning in the Fisher Information Matrix, resulting in slower convergence with stochastic gradient descent. However, using natural gradient descent or appropriate preprocessing techniques can mitigate these issues, allowing for similar convergence across both encoding types. The findings offer practical guidance on selecting variable encodings and preprocessing methods for Boltzmann machines. AI

IMPACT Provides practical guidelines for optimizing Boltzmann machine learning performance through variable encoding choices.

RANK_REASON Academic paper detailing research findings on machine learning encodings. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• Boltzmann Machine Learning Using Mean Field Theory and Linear Response Correction
• Fisher Information Matrix
• Ising
• Natural Gradient Descent
• QUBO
• stochastic gradient descent
• Yasushi Hasegawa

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →