1
$\begingroup$

I am reading this book called "Deep Learning" (by Goodfellow, Bengio and Courville).

On page 326, in the first paragraph, it says:

CNNs, are a specialized kind of neural network for processing data that has a known grid-like topology. Examples include time-series data, which can be thought of as a 1-D grid taking samples at regular time intervals, and image data, which can be thought of as a 2-D grid of pixel

Considering an image as a grid is completely intuitive. And, similarly, we can extend the logic to a 1-D time series.

But then what cannot be considered as having a grid-like structure?

$\endgroup$
1
$\begingroup$

Graphs.

In graphs, nodes and edges do not usually have a specified position/orientation: for example, we cannot say that node $A$ is to the right/left of node $B$ because edges do not typically have an orientation. In grid-like structures, like images, we could say that about pixels. In a graph, all we know is that nodes are connected to other nodes via some edges, which may have weights, but, again, no orientation.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.