Dense networks

• dense networks are fully connected. Each node is connected to each input and output node
• dense networks are used on top of other networks to improve accuracy

Convolution networks

• These are best known for machine vision applications
• object recognition of self driving cars, face recognition, classify images etc
• They can analyze any data with spatial pattern. Even NLP has spatial pattern which convolution network can work on

Recurrent neural network (RNN)

• used for data that come in a sequence
• NLP data, financial data etc
• LSTM -> specialized RNN Long short term memory network
• beyond scope

Reinforcement learning

• this network gives feedback on its performance
• defeat player in go
• beyond scope

Generative adversarial network

• make complex images very realistic
• take picture and get it painted as if done by picasso
• beyond scope