recurrent neural network

Terms from Artificial Intelligence: humans at the heart of algorithms

Page numbers are for draft copy at present; they will be replaced with correct numbers when final book is formatted. Chapter numbers are correct and will not change now.

A recurrent neural network is a form of neural network with a state/memory that runs on time series or sequential data. At each step a neura networ uses the current state and inout token to geerate the next state and to predict the next token. As the hidden state is not part of the training data, the training phase is more complex than for a simple neural network.

Defined on page 331

Used on Chap. 14: pages 331, 332, 338

Also known as RNN

Recurrent neural network (RNN)

Running a recurrent neural network on the data stream d1, d2, d3, ...