Long Short-Term Memory (LSTM) is widely used in different areas of deep learning. It is developed using a neural network with recursive architecture; that’s why it also maintains a feedback connection. LSTM can process singular data units (such as images) and sequential data units (such as speech or video), for example, speech recognition and handwriting. It consists of a cell, an input gate, an output gate, and a forget gate. Read more

May, 1997