Hacker News Reader

Just based on the docs, on the lowest level they both create a mapping from input features to relations. Then NeuraLogic has an example for implementing GCN like message passing with a simple relational rule:

``` node2(X) <= W node1(Y), edge(X, Y) ```

Do you know, is it possible to implement something on Scallop as well, or are the differences much larger?

View on HN

The GCN implementation is based on the concept of templating (or lifting) which is not exploited in Scallop. In fact, this is the key idea of Neuralogic for joining deep learning and relational-logic representations.

It is explained in the paper "Beyond Graph Neural Networks with Lifted Relational Neural Networks" (https://arxiv.org/abs/2007.06286) and you also have a series of blog posts at https://medium.com/@sir.gustav

Reply on news.ycombinator.com