NeutronStream is a Dynamic GNN Training Framework with Sliding Window for Graph Streams. NeutronStream abstracts the input dynamic graph into a chronologically updated stream of events and processes the stream with an optimized sliding window to incrementally capture the spatial-temporal dependencies of events. In addition, NeutronStream provides a parallel execution engine to tackle the sequential event processing challenge to achieve high performance. NeutronStream also integrates a built-in graph storage structure that supports dynamic updates and provides a set of easy-to-use APIs that allow users to express their dynamic GNNs.
- glog-0.60
- gtest
- NUMCPP
- libtorch-cu
First clone the repository and initialize the submodule:
git clone https://github.com/superccy/NeutronStream.git
cd NeutronStream
git submodule update --init --recursive
# or just use one command
git clone --recurse-submodules https://github.com/superccy/NeutronStream.gitTo build:
mkdir build && cd build
cmake ..
make All datasets we used:
| Datasets | Nodes | Edge.init | Edge.final | Event.num |
|---|---|---|---|---|
| Social Evolution | 84 | 575 | 794 | 54,369 |
| Github | 284 | 298 | 4,131 | 20,726 |
| DNC | 2,029 | 0 | 5,598 | 39,264 |
| UCI | 1,899 | 0 | 20,296 | 59,835 |
| Reality | 6,809 | 0 | 9,484 | 52,052 |
| Slashdot | 51,083 | 0 | 131,175 | 140,778 |