v0.10.0

GeometricFlux v0.10.0

Diff since v0.9.0

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

Closed issues:

• examples/gde.jl not working (#264)
• This package does not respect semantic versioning (#266)

Merged pull requests:

• Fix neural GDE example (#268) (@yuehhua)
• Fix GAE example (#269) (@yuehhua)
• Update docs (#271) (@yuehhua)

v0.9.0

Full Changelog: v0.8.0...v0.9.0

v0.8.2

What's Changed

• Create flux.css to style the docs by @logankilpatrick in #222
• Implement CGConv layer. by @emsal0 in #220
• Add AbstractGraphLayer for layers that accept graphs by @yuehhua in #205
• Append more info to the GNN construction docs by @emsal0 in #229
• Resolve gradient bug for GatedGraphConv by @yuehhua in #250
• Correct GCNConv with normalized_adjacency_matrix by @yuehhua in #234
• Update GCN model by @yuehhua in #251

New Contributors

• @logankilpatrick made their first contribution in #222

Full Changelog: v0.7.7...v0.8.2

v0.8.0

GeometricFlux v0.8.0

Diff since v0.7.7

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

Closed issues:

• Implementation of neural operators (#172)
• move FeaturedGraph to COO format (#200)
• Stable docs 404 in README (#221)
• examples/gcn.jl doesn't work (#245)
• the GCNConv implementation is not correct (#249)

Merged pull requests:

• Add AbstractGraphLayer for layers that accept graphs (#205) (@yuehhua)
• Implement CGConv layer. (#220) (@emsal0)
• Create flux.css to style the docs (#222) (@logankilpatrick)
• Append more info to the GNN construction docs (#229) (@emsal0)
• Correct GCNConv with normalized_adjacency_matrix (#234) (@yuehhua)
• Resolve gradient bug for GatedGraphConv (#250) (@yuehhua)
• Update GCN model (#251) (@yuehhua)

v0.7.7

GeometricFlux v0.7.7

Diff since v0.7.6

• 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

Closed issues:

• Increase test cases (#7)
• Method error when training GCNConv with FeaturedGraph input (#76)
• More "mutating arrays is not supported" (#103)
• missing adjoints for pooling functions with specified graph size (#147)
• drop ScatterNNlib in favor of NNlib.scatter (#191)
• drop support in layer for arbitrary graph types and use only FeaturedGraphs (#201)
• FeaturedGraph cannot be moved to gpu (#202)
• deprecate FeatureSelector (#207)
• broken documentation links (#210)

Merged pull requests:

• CompatHelper: bump compat for "CUDA" to "3.0" (#175) (@github-actions[bot])
• Use mapreduce in message-passing (#179) (@yuehhua)
• Refactor pool function into scatter function (#180) (@yuehhua)
• Update doc (#181) (@yuehhua)
• Fix building doc (#182) (@yuehhua)
• Some cleanup (#183) (@yuehhua)
• Update docs (#184) (@yuehhua)
• Implement GINConv layer. (#186) (@emsal0)
• Replace ScatterNNlib with NNlib (#187) (@yuehhua)
• Refactor (#189) (@yuehhua)
• remove learnable parameters when bias=false (#196) (@CarloLucibello)
• misc cleanup and add buildkite (#198) (@CarloLucibello)
• homogenize forward pass for all layers (#203) (@CarloLucibello)
• misc improvements (#206) (@CarloLucibello)
• Improve documentation (#208) (@yuehhua)
• Fix doc (#211) (@yuehhua)
• add benchmark script (#214) (@CarloLucibello)
• Fix GINConv docstring. (#219) (@emsal0)

v0.7.6

GeometricFlux v0.7.6

Diff since v0.7.5

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

Closed issues:

• Consider to move into the Flux organisation (#2)
• could you please add a reference for the notions you used here (and also in other places) ? (#164)
• Tag Flux.jl v0.12 compat (#165)
• Conflict between Flux and ScatterNNLib inside GeometricFlux (#169)

Merged pull requests:

• Add dimensional check for each layer (#143) (@yuehhua)
• Support Zygote v0.6 (#145) (@yuehhua)
• CompatHelper: bump compat for "Reexport" to "1.0" (#146) (@github-actions[bot])
• Add google analytics (#151) (@yuehhua)
• Bump to CUDA v2.5 and FillArrays v0.11 (#152) (@yuehhua)
• Add JuliaCon paper (#155) (@yuehhua)
• Update figures (#156) (@yuehhua)
• Try to deploy stable doc (#157) (@yuehhua)
• Refine paper figures (#158) (@yuehhua)
• Correct reference (#160) (@yuehhua)
• Add paper.pdf (#161) (@yuehhua)
• Update README.md (#162) (@CarloLucibello)
• Add ScatterNNlib namespace for gather and scatter methods (#170) (@ilancoulon)
• Fix bugs while updating to Julia v1.6 and Flux v0.12 (#171) (@yuehhua)

v0.7.5

GeometricFlux v0.7.5

Diff since v0.7.4

• 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

Closed issues:

• Tag? (#99)
• Cannot give transposed features to GCNConv (#110)
• GCN example still doesn't work (#125)
• GATConv aggregates by default (#129)
• GATConv does not seem to have the desired behaviour (#130)

Merged pull requests:

• Improve differentiability test (#121) (@yuehhua)
• Refactor propagate into a generic version (#122) (@yuehhua)
• Refactor graph convolutional layers (#123) (@yuehhua)
• Fix GCN on CPU by converting into Matrix (#126) (@yuehhua)
• Fix GATConv computation of concat=false (#127) (@yuehhua)
• Making Conv layers accept transposed matrices (#128) (@ilancoulon)
• Fix softmaxing of alphas in GATConv (#131) (@ilancoulon)
• CompatHelper: bump compat for "CUDA" to "2.2" (#132) (@github-actions[bot])
• Refactor dispatching in GATConv (#133) (@yuehhua)
• match matrix type to prevent CUDA.CUSPARSE error (#134) (@alperyilmaz)
• Remove deps StaticArrays (#137) (@yuehhua)
• Correct type of negative slope in leaky relu again (#141) (@yuehhua)

v0.7.4

GeometricFlux v0.7.4

Diff since v0.7.3

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

Merged pull requests:

• Support CUDA v2.1 (#118) (@yuehhua)
• Refactor (#119) (@yuehhua)

v0.7.3

GeometricFlux v0.7.3

Diff since v0.7.2

• 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

Merged pull requests:

• Add activation function for GraphConv (#109) (@yuehhua)
• onecold() should be performed at cpu (#112) (@alperyilmaz)
• cache argument triggers error (#114) (@alperyilmaz)
• fix typos in multiple files (#115) (@alperyilmaz)
• Some new features (#116) (@yuehhua)

v0.7.2

GeometricFlux v0.7.2

Diff since v0.7.1

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

Closed issues:

• Cannot differentiate GATConv (#54)

Merged pull requests:

• Make message passing layers differentiable + test that it can be computed (#102) (@ilancoulon)
• Doc (#105) (@yuehhua)
• Bump CUDA to v2.0 (#107) (@yuehhua)
• Refactor (#108) (@yuehhua)