Deep Graph Infomax (DGI) is a method for learning node representations in graph-structured data without supervision. DGI maximizes mutual information between local node representations (patches) and global graph summaries, using graph convolutional networks. It does not rely on random walk objectives and is applicable to both transductive and inductive learning. DGI learns patch representations that summarize subgraphs around nodes, which can be used for downstream tasks. The method uses a discriminator to maximize mutual information between local and global representations, with a binary cross-entropy loss. DGI achieves competitive performance on node classification benchmarks, sometimes outperforming supervised learning. The method is evaluated on various datasets, including Cora, Citeseer, Pubmed, Reddit, and PPI networks. DGI outperforms unsupervised GraphSAGE approaches and performs competitively with supervised methods on some datasets. Qualitative analysis shows that DGI embeddings cluster well and capture structural properties of graphs. DGI is effective in both transductive and inductive settings, and its performance is validated through extensive experiments and comparisons with other methods.Deep Graph Infomax (DGI) is a method for learning node representations in graph-structured data without supervision. DGI maximizes mutual information between local node representations (patches) and global graph summaries, using graph convolutional networks. It does not rely on random walk objectives and is applicable to both transductive and inductive learning. DGI learns patch representations that summarize subgraphs around nodes, which can be used for downstream tasks. The method uses a discriminator to maximize mutual information between local and global representations, with a binary cross-entropy loss. DGI achieves competitive performance on node classification benchmarks, sometimes outperforming supervised learning. The method is evaluated on various datasets, including Cora, Citeseer, Pubmed, Reddit, and PPI networks. DGI outperforms unsupervised GraphSAGE approaches and performs competitively with supervised methods on some datasets. Qualitative analysis shows that DGI embeddings cluster well and capture structural properties of graphs. DGI is effective in both transductive and inductive settings, and its performance is validated through extensive experiments and comparisons with other methods.