Seminars

Within the machine learning community, there has been a  

growing interest in learning structured models from input data that  

is itself structured.  Graph identification refers to methods that  

transform an observed input graph into an inferred output graph.  

Examples include inferring organizational hierarchies from social  

network data and identifying gene regulatory networks from protein- 

protein interactions.  Graph alignment refers to a related problem,  

the problem of matching nodes, edges, and larger subgraphs between  

two graphs.  The key processes in graph identification and alignment  

are: entity resolution, link prediction, and collective  

classification.  I will overview algorithms for these tasks, discuss  

the need for integrating the results to solve the overall problem  

collectively, and show how these methods are relevant to foundational  

problems in AI such as knowledge representation, reformulation, and  

reasoning.