LGDADA Official Trailer: Shall we cooperate but in a competitive manner?

#paperTrailer #graphSynthesis #GCN #networkNeuroscience This is the #story of #LGDADA deep learning architecture. 🤔 Question: How can we morph a source graph of type A nested in a heterogeneous source domain into a target graph of type B? Sounds pretty abstract🤔? Check the video for a simpler formalization 🤩. 🤩 Answer: Our recent #MEDIA2021 journal paper (IF: 11.28) first-authored by Alaa Bessadok pushes the boundary of graph neural networks to cross-domain brain graph synthesis, a fractionally explored field. The proposed Learning-guided Graph Dual Adversarial Domain Alignment (LGDADA) architecture generates a target graph B from an input source graph A using geometric deep learning (#GDL) and generative adversarial networks (#GAN). LGDADA is trained using a *dual adversarial regularization* approach. The learning process integrates the target graph prediction and the domain alignment in a single unified adversarial framework where both tasks cooperate —hence the term #cooperatively #adversarial. For more, check out: ** the LGDADA paper: https://lnkd.in/ehYr_dU ** the LGDADA GitHub code: https://lnkd.in/e_6Y3ds Soundtrack: the Lion King 🦁