The success of deep learning methods has enabled significant progress in some areas of AI, but has been mostly limited to reactive systems that require large amounts of annotated data. There is an increasing interest in planning from the machine learning community. However, an important gap exists between both communities in terms of objectives, techniques, or methodologies. For example, learning-based systems tend to scale poorly on problems of combinatorial nature, in which planning-based systems excel, and similarly, planning-based systems have very limited generalization capabilities, which is a fundamental quality of learning-based systems. The Planning and Learning track aims to bridge this gap by providing an opportunity for both communities to engage directly. In particular, we are interested in how the progress we have seen from the deep learning side can be integrated into the progress we have seen in planning.
This track has already become an important part of the main ICAPS conference and in this fourth edition we welcome both theoretical and empirical work addressing (but not limited to) the following topics:
- Planning and efficient exploration in reinforcement learning
- Planning and representation learning
- Planning and relational learning
- Learning to improve the effectiveness of planning & scheduling systems
- Planning & scheduling in learned domain models
- Learning effective heuristics and other forms of control knowledge
- Planning applied to automating machine learning systems
- Learning of cost functions used in planning
- November 15, 2019 Abstracts due (electronic submission)
- November 20, 2019 Papers due (electronic submission, PDF)
- January 6-9, 2020 Author feedback period
- January 20, 2020 Notification of acceptance or rejection
The reference time-zone for all deadlines is UTC-12. That is, as long as there is still some place anywhere in the world where the deadline has not yet passed, you are on time!