经验总结

  1. 首页 > 生活 > >

GNN 101( 三 )

生活百科GNN 101


Scatter-Gather

  • Fuse adjacent graphs ops
    One common fuse pattern for GCN & GraphSAGE：
    GNN 101

    文章插图
    Challenge: How to fuse more GNN patterns on different ApplyEdge and ApplyVertex,automatically?
  • How to implement fused Aggregate
    GNN 101

    文章插图
    Challenge:
    • Different graph data structureslead to different implementations in same logic operations;
    • Different graph characteristics favors different data structures;(like low-degree graphs favor COO, high-degree graphs favor CSR)
    • How to find the applicable zone for each and hide such complexity to data scientists?
More
  • Inference challenge
    • GNN inference needs full batch inference, how to make it efficient?
    • Distributed inference for big graph?
    • Vector quantization for node and edge features?
    • GNN distilled to MLP?
  • SW-HW co-design challenge
    • How to relief irregular memory access in scatter-gather?
    • Do we need some data flow engine for acceleration?
  • …
Finishing words“There is plenty of room at the top” 对技术人员很重要 。但为避免入宝山而空返 ， 我们更需要建立起技术架构 ， 这就像是地图一样 ， 只有按图索骥才能更好地探索和利用好top里的plenty of room 。
GNN 101

文章插图
References
  1. Graph + AI: What’s Next? Progress in Democratizing Graph for All
  2. Recent Advances in Efficient and Scalable Graph Neural Networks
  3. Crossing the Chasm – Technology adoption lifecycle
  4. Understanding and Bridging the Gaps in Current GNN Performance Optimizations
  5. Automatic Generation of High-Performance Inference Kernels for Graph Neural Networks on Multi-Core Systems
  6. Understanding GNN Computational Graph: A Coordinated Computation, IO, And Memory Perspective
  7. Graphiler: A Compiler For Graph Neural Networks
  8. Scatter-Add in Data Parallel Architectures
  9. fuseGNN: Accelerating Graph Convolutional Neural Network Training on GPGPU
  10. VQ-GNN: A Universal Framework to Scale up Graph Neural Networks using Vector Quantization
  11. NeuGraph: Parallel Deep Neural Network Computation on Large Graphs
  12. Completing a member knowledge graph with Graph Neural Networks
  13. PinnerFormer: Sequence Modeling for User Representation at Pinterest
  14. Gartner and Graph Analytics

经验总结扩展阅读


上一篇：2023年9月14日坐船好吗 2023年9月14日坐船黄道吉日

下一篇：哪一种芦荟胶比较好用 十款性价比高的芦荟胶排行榜