NetInf 2017 - Workshop on Inferring Networks from Non-Network Data (NetInf17)

Network representation learning is an emerging field that is exclusively focused on developing the understanding and rigor for learning and inferring useful network representations from noisy, indirect and diverse data measurements. These data are prevalent in science and industry in the form of genetic expression, brain imaging, geotagged and geo-social data, and user behavioral/interaction data on the web.
Researchers aim to learn a network representation of this data because a network is a convenient and powerful model for investigating the system or population in question. However, researchers are often faced with having to make arbitrary decisions on how to construct networks from underlying data. Such arbitrary decisions and the ad-hoc representations they lead to have important implications on the performance of subsequent learning tasks. There is a need for novel, rigorous methods in network construction and validation, and the challenges and pitfalls along the network science pipeline, from underlying data to domain investigation.
This workshop will explore recent developments in the area of network representation learning including:
Multi-modal and heterogeneous techniques that lead to robust network representations
Strategic and adoptive data collection mechanisms for discovering the network
Task-oriented network representation learning
Validation of network representations in the absence of ground truth
This area is related to several communities working in broader network science, including dynamic networks, parametric network modeling and relational learning and statistical inference on networks.