October 2018
Beginner
362 pages
9h 32m
English
A Neural Tensor Network (NTN) is a new form of neural network that works like a standard feed-forward network, only it contains something known as a tensor layer instead of standard hidden layers. The network was originally developed as a means of completing knowledge bases by connecting unconnected entities. For instance, if we had the entities Google and YouTube, the network would help connect the two entities so that Google -> Owns -> YouTube. It passes different relationship pairs through the network instead of through a singular vector, and it does this by passing them as a tensor. Each slice of that tensor represents a different variation of a relationship that two entities can have.
In the ...
Read now
Unlock full access