Skip to content

Latest commit

 

History

History
279 lines (196 loc) · 6.01 KB

CHANGELOG.md

File metadata and controls

279 lines (196 loc) · 6.01 KB
Raw

Changelog

All notable changes to this project will be documented in this file.

[0.14.0]

  • update compat
  • support Julia greater than or equals to v1.9

[0.13.10]

  • update compat

[0.13.9]

  • update compat

[0.13.8]

  • adapt to new NNlib API within_gradient

[0.13.7]

  • fix GatedGraphConv

[0.13.6]

  • fix tests
  • update doc

[0.13.5]

  • replace GraphMLDatasets in favor of MLDatasets
  • update examples and docs

[0.13.4]

  • support GraphSignals to 0.7

[0.13.3]

  • update doc for FeaturedGraph

[0.13.2]

  • fix doc

[0.13.1]

  • GraphParallel support positional_layer

[0.13.0]

  • implement EEquivGraphConv layer with nested design
  • support positional encoding from GraphSignals
  • add LSPE

[0.12.4]

  • replace ADAM as Adam

[0.12.3]

  • update doc for graph network

[0.12.2]

  • replace Zygote.ignore as ChainRulesCore.ignore_derivatives

[0.12.1]

  • remove @deprecate

[0.12.0]

  • add roadmap
  • add SAGEConv layer
  • support dynamic graph update

[0.11.1]

  • fix link

[0.11.0]

  • Adds GATv2 layer
  • add DeepSet model and digit sum example
  • fix GAT example
  • add tutorials
  • replace Flux.destructure by Optimisers.destructure

[0.10.1]

  • fix VGAE example and correct precision
  • implement new message-passing scheme, including GraphConv, GATConv, GatedGraphConv, EdgeConv, GINConv and CGConv layers
  • fix tests for GraphNet
  • add WithGraph for Chain

[0.10.0]

  • update docs and add defining GNN layer to doc
  • update GAE example
  • fix neural GDE example

[0.9.0]

  • add semisupervised gcn and gcn with fixed graph example
  • implement new GCNConv
  • add node2vec
  • bug fix

[0.8.0]

  • correct GCNConv with normalized_adjacency_matrix
  • add L2 regularization to gcn example
  • migrate Graphs, GraphSignals, GraphLaplacians and examples
  • resolve gradient bug for GatedGraphConv

[0.7.7]

  • drop support of julia v1.4 and v1.5
  • support CUDA v3.3
  • support Flux v0.12
  • fix stable doc
  • add benchmark script
  • migrate scatter to NNlib
  • make gradient of GatedGraphConv available
  • Implement GINConv layer. (#186)
  • check consistency for vertex or edge number between graph and features
  • add manual for pooling layers and bypass_graph
  • deprecate FeatureSelector
  • not export GraphNetwork and MessagePassing APIs
  • new implementation for message-passing scheme

[0.7.6]

  • Add dimensional check for each layer
  • Support Flux up to v0.12
  • Support CUDA up to v2.6
  • Support Zygote up to v0.6

[0.7.5]

  • FeaturedGraph API change
  • Refactor graph net and message passing framework
  • Improve differentiability test
  • Refactor GCNConv and ChebConv operator
  • Fix bug in GATConv layer
  • Update GAT example
  • Cast testing data to Float32
  • Support CUDA up to v2.2
  • Support transpose input of a layer
  • Replace Travis CI by Github Action CI

[0.7.4]

  • Adjust edge_index_table API for directed
  • apply_batch_message as API
  • Support CUDA v2.1
  • Refactor
  • Fix bug

[0.7.3]

  • Add bypass_graph
  • Support FeaturedGraph as input graph for GCNConv
  • Add node index for message/update function
  • Add activation function for GraphConv
  • Reexport GraphSignals
  • Support FillArrays v0.10
  • Bug fix

[0.7.2]

  • Differentiability test
  • Refactor GN for differentiability
  • Remove cache argument from layer
  • Add docs
  • Bump CUDA to v2.0
  • Add paper

[0.7.1]

  • Add GraphMLDatasets as dependency to provide datasets

[0.7.0]

  • VGAE example available
  • Add Planetoid and Cora dataset

[0.6.3]

  • GDE, GAE VGAE examples available
  • Correct GCNConv show

[0.6.2]

  • Add FeatureSelector
  • Correct ChebConv computation
  • Make scaled_laplacian differentiable
  • Add ScatterNNlib and GraphSignals as deps
  • Improve GAT example
  • Upgrade to CUDA
  • Maintain Travis CI

[0.6.1]

  • Update to CUDA 1.2 and Flux 0.11
  • Refactor graph-related API
  • Improve learning rate in example

[0.6.0]

  • Rewrite graph network GraphNet and message passing MessagePassing framework
  • Expand functionality of FeaturedGraph to support node_feature, edge_feature and global_feature
  • Speed up ChebConv layer
  • Speed up scatter functions
  • Add graph index-related functions
  • GCN example works and increase training stablility
  • Fix show GCNConv
  • Add more test for linear algebra
  • Update cpu scatter benchmark plot and scripts

[0.5.2]

  • Add scaled Laplacian
  • Support CuArrays v2.0 and Flux v0.10.4
  • ChebConv, GraphConv, GATConv, GatedGraphConv and EdgeConv support FeaturedGraph
  • Add SimpleWeightedGraphs and MetaGraphs as deps
  • Fix broadcastly casting error

[0.5.1]

  • GCNConv layer supports FeaturedGraph (#34)
  • Support linear algebra for FeaturedGraph
  • Add nv API for FeaturedGraph
  • Add LightGraphs as dependency
  • Correct normalized laplacian type
  • Fix bug in normalized_laplacian
  • Fix Base.show on GCNConv
  • Add docs (#35)

[0.5.0]

  • Support scatter operations for MArray (#32)
  • Support GCNConv layer accepting graph input (#31)

[0.4.0]

  • Compatible with Julia v1.4 while not support before v1.3
  • Not support old version CuArrays, CUDAnative and CUDAapi
  • Improve performance of scatter operations for CPU and new benchmark (#29)
  • Scatters support almost all Real numbers except Bool on CPU
  • Add benchmark for scatter operations
  • Implement TopKPool layer (#22)

[0.3.0]

  • Improve performance of scatter operations in both CPU/CUDA version
  • Add benchmark result
  • Add multihead GAT on graph support
  • Move pool_dim_check to Dims constructor

[0.2.0]

  • Available on Julia v1.2 and v1.3
  • Convolution layers works with CUDA
  • Provide scatter add, sub, mul, div, max, min, mean for CPU and CUDA
  • Provide pool add, sub, mul, div, max, min, mean for CPU and CUDA
  • Provide gradient of scatter add, sub, mul, div, max, min, mean for CPU and CUDA
  • Provide gradient of pool add, sub, mul, div, max, min, mean for CPU and CUDA
  • Provide gather
  • Provide good abstract for graph network block
  • Integrate message passing scheme and graph network block
  • Add logo
  • Add docs
  • Add layer docs and Base.show
  • Provide dynamically change graph in runtime
  • Provide GlobalPool layer