تقرير
EA-CG: An Approximate Second-Order Method for Training Fully-Connected Neural Networks
العنوان: | EA-CG: An Approximate Second-Order Method for Training Fully-Connected Neural Networks |
---|---|
المؤلفون: | Chen, Sheng-Wei, Chou, Chun-Nan, Chang, Edward Y. |
سنة النشر: | 2018 |
المجموعة: | Computer Science |
مصطلحات موضوعية: | Computer Science - Machine Learning |
الوصف: | For training fully-connected neural networks (FCNNs), we propose a practical approximate second-order method including: 1) an approximation of the Hessian matrix and 2) a conjugate gradient (CG) based method. Our proposed approximate Hessian matrix is memory-efficient and can be applied to any FCNNs where the activation and criterion functions are twice differentiable. We devise a CG-based method incorporating one-rank approximation to derive Newton directions for training FCNNs, which significantly reduces both space and time complexity. This CG-based method can be employed to solve any linear equation where the coefficient matrix is Kronecker-factored, symmetric and positive definite. Empirical studies show the efficacy and efficiency of our proposed method. Comment: Change to AAAI-19 Version |
نوع الوثيقة: | Working Paper |
الوصول الحر: | http://arxiv.org/abs/1802.06502Test |
رقم الانضمام: | edsarx.1802.06502 |
قاعدة البيانات: | arXiv |
الوصف غير متاح. |