Recurrent Neural Networks

• Sequential models
• This allows us to classify or label sequential inputs, generate sequences of text or translate one sequence to another
• It provvides building blocks that fit well into the toolkit of fully connected convolutional layers.

Introduction

• Audio & video clips

• sequence of characters

• stock market prices

• RNN are family of networks that explicitly model time

• RNN bild on the same neurons summing up weighted inputs from other neurons

• Neurons are allowed to connect both forward & higher layers & backward to lower layers andd form cycles.

• RNN are basically directed graphs of neurons & weights

• Input neurons have on incomming connections because their data

Back Propogation Through Time