Hi,
Thank you for developing this wonderful tool. I would like to implement a user-defined aggregation function with learnable parameters. Now I need to feed the input graph to my user-defined message function and reduce function as an input parameter, something like graph.update_all(self.message_func, self.reduce_func(graph)). Is it possible to achieve this goal with DGL? I would really appreciate the help. Thank you.
Best,
Yongcheng