SGD Converges to Global Minimum in Deep Learning via Star-convex Path.
Yi ZhouJunjie YangHuishuai ZhangYingbin LiangVahid TarokhPublished in: CoRR (2019)
Keyphrases
- global minimum
- deep learning
- cost function
- global minimization
- energy function
- simulated annealing
- energy minimization
- unsupervised learning
- global minimizers
- unsupervised feature learning
- global optimum
- machine learning
- global minima
- mental models
- weakly supervised
- energy functional
- active contours
- supervised learning
- input image
- prior knowledge
- object recognition
- objective function
- bayesian networks