home

 

Long short-term memory (LSTM)

Long short-term memory units (or blocks) are a building unit for layers of a recurrent neural network (RNN).

Nominated by Aschwin Vijayakumar (intel):

“They help machines understand the context, as against trying to process images or audio in a instance by instance update. For example if you’re trying to process a video for a smart home security camera. Me and my wife would like the system to ‘keep an eye’ on our toddler. LSTM can understand context, it can detect if my son is about to jump off the table, if he is running around, jumping on the couch. You cannot do these things with something that doesn’t have memory like a LSTM.

It is necessary to help bring vidon processing to the next stage.”

Aschwin Vijayakumar


Social tagging: