Graph-based learning techniques have seen a wide range of applications in machine learning. Many forms of data are naturally modeled as a graph, such as networks of social media users, databases of images, states of large physical and biological systems, or collections of DNA sequences. Graph structure encodes interdependencies among constituents and provides for a convenient representation of the high-dimensional data. Patterns in the graph structure can be used to inform the design of scalable, graph-based learning algorithms in semi-supervised and unsupervised settings that deal with a limited amount of labeled information.
This workshop will explore recent theoretical advances in graph-based learning and the opportunities they present for improving the performance and scalability of algorithms while reducing the cost associated with labeling data.