what is the main diference between SSC(Submanifold sparse convolutional networks) and SRR(Sparse Recursive Representations)?

[978326187]

after reading the paper, i think SRR is an extension of SSC in event data, can anyone explain that?

[978326187]

i think the difference is that asynet propagate event-by-event.

[MessikommerNico]

That is correct. The difference between SSC and asynet is the possibility to propagate single events through the network. Due to the rule book formulation and the network state propagation through time, asynet reduces the number of FLOPS required to process single events or small batches of events. In comparison, SCC compute the activations of active sites even if the input did not change. This is not the case for asynet due to the network state propagation through time.