Graph, Data-science, and Deep Learning
0 likes2,007 views
The document discusses graph-based deep learning, emphasizing its advantages over traditional matrix-based methods for handling natural language processing (NLP) tasks. It highlights the importance of time in analysis and proposes a focus on graph visualization rather than on software complexities. Key contributors to the field, such as Richard Socher and Andrew Ng, are mentioned, along with a link to a demonstration related to sentiment analysis.
Graph, Data-science, and Deep Learning
- 1. Wadda you mean, Graph?Wadda you mean Graph?
- 5. A graph is a graph is a graph what drugs will bind to protein X and not interact with drug Y?
- 6. Graph-based Deep Learning • Graphs + NLP = AWESOME • Novel possibilities: Time is a fundamental factor for any analysis • Graph (natural NLP and other fancy affinities) vs Matrices (complicated, not scalable algorithms/ mathematics) • Whiteboard/Visualization friendly • Thin Application layer = Focus on the graph, not the software layer
- 7. Richard Socher, Alex Perelygin, Jean Wu, Jason Chuang, Chris Manning, Andrew Ng and Chris Potts. 2013. Recursive Deep Models for Semantic Compositionality Over a Sentiment Treebank. EMNLP 2013 code & demo: http://nlp.stanford.edu/sentiment/index.html Graph-based Deep Learning*