Modular deep belief networks that do not forget

Computer Science

• Learned Feature 100%
• High Dimensional Data 50%
• Input Distribution 50%
• Feature Mapping 50%
• handwritten digit 50%
• Compact Representation 50%
• Learning Algorithm 50%
• Learning Rate 50%
• Training Data 50%
• Modularization 50%

Psychology

• Learning Algorithm 100%
• Training Set 100%

Keyphrases

• Feature Mapping 12%