Backpropagation

Compute interpretation

Credit-assignment mechanism for layered neural networks; its historical role depends on when data, model size, memory, and hardware made deeper networks trainable.

Supporting reading cards

• Learning representations by back-propagating errors (1986, pre_2012_cpu_statistical_foundations)
• Gradient-based learning applied to document recognition (1998, pre_2012_cpu_statistical_foundations)
• A fast learning algorithm for deep belief nets (2006, pre_2012_cpu_statistical_foundations)
• Reducing the dimensionality of data with neural networks (2006, pre_2012_cpu_statistical_foundations)

Obsolete or less central under later compute

Track this only through linked reading cards; do not treat this method page as standalone evidence.