博文
Coursera: Neural Networks for ML- Lecture 14
|||
DBN(信息量较大)
2006_Hinton_A fast learning algorithm for deep belief nets
2006_Science_Reducing the Dimensionality of data with neural networks
2009_The difficulty of training deep architectures and the effect of unsupervised pre-training
ICML2010_Rectified Linear Units Improve Restricted Boltzmann Machines
2. Discriminative fine-tuning for DBNs
3. What happens during discriminative fine-tuning?
这一节主要讲了为什么unsupervised pre-training对训练多层结构有帮助，其中的阐述和实验结果具体可以参见这两篇文章：
Erhan_2009_The difficulty of training deep architectures and the effect of unsupervised pre-training
Erhan_2010_Why does unsupervised pre-training help deep learning?
4. Modeling real-valued data with an RBM
5. RBMs are infinite Sigmoid Belief Nets(Optional)
https://blog.sciencenet.cn/blog-1583812-820528.html
上一篇：Coursera: Neural Networks for ML- Lecture 13
下一篇：Ubuntu+CUDA6.5+Caffe安装配置汇总
扫一扫，分享此博文