2020
Applications
Graph-based learning
in practice.
Using relational structure to model language, dynamic scenes, and decisions in unknown networks.
Dialogue response generation
“On Incorporating Structural Information to Improve Dialogue Response Generation” explores structural information for generating dialogue responses. It was presented at the NLP for Conversational AI Workshop at ACL 2020, following an earlier extended abstract at EurNLP 2019.
Dynamic scenes and vehicle behavior
Collaborative work on understanding dynamic scenes uses graph convolution networks to model relationships in scene data. Related work studies vehicle behavior classification with multi-relational graph convolutional networks.
Influence maximization in unknown networks
Our AAMAS 2020 work studies learning policies for effective graph sampling in unknown social networks, in the context of influence maximization. It connects reinforcement learning with graph-based decision-making and was nominated for a Best Paper Award.
Related publications & artifacts
2020
2020
2020