Keras 2.x Projects

Multilayer neural networks

The term multilayer neural networks denotes neural networks composed of many hidden levels (at least two) organized hierarchically. Hierarchical organization allows you to share and reuse information. Along the hierarchy, you can select specific features and discard unnecessary details in order to maximize the invariance. In multilevel machine learning, the deeper levels take inputs from the outputs of previous layers and perform more transformations and abstractions on them. This management of learning levels is inspired by the way in which a mammalian brain processes information and learns, responding to external stimuli. The following diagram shows a generic architecture of a multilayer neural network (with two ...