Recently, geometric deep learning is a popular field in deep learning. It can process unstructural features on graph, which is different from traditional data such as images or text. It can further be applied to social network, recommender systems, molecular structure and biological network. I will introduce a geometric deep learning framework GeometricFlux implemented in Julia.
Many research data with intrinsic structure which lies in non-Euclidean space. Geometric deep learning plays a role in modeling non-Euclidean data with graph structure. I introduce GeometricFlux, a Julia package for geometric deep learning on graph. GeometricFlux relies on Zygote as automatic differentiation engine, accepts graph data structure provided by JuliaGraph. GeometricFlux layers are compatible with Flux layers and supported by CuArrays. It will be a competitive platform against other framework.
About Yueh-Hua Tu
PhD. candidate in TIGP bioinformatics, Academia sinica. He is the host of Julia Taiwan User Group community. He has background of medical laboratory science and computer science. He has been machine learning lecturer in ITRI in Taiwan and now active in machine learning/deep learning communities in Taipei and Taichung. He is also an open source contributor and maintainer of GeometricFlux.jl. He has two books of Julia programming and Julia data science and scientific computing in traditional Chinese.