Our group at IBM Research in Thomas J. Watson Research Center is looking for candidates for the summer 2021 internship. The general description of the project can be found below.
If you are interested, please reach out to me at michael.katz1@ibm.com as soon as possible.
Project description:
Sequential decision making is one of the most important problems in modern AI. Two main subfields of AI that deal with sequential decision making are Reinforcement Learning (RL) and AI Planning. Each of these approaches has their strong sides and their weaknesses. AI Planning is a model-based approach, relying on a well-specified symbolic model to guide the search for a solution. It does not require additional data beyond the symbolic model, domain-independent, being completely agnostic to the problem behind the model and allowing for solving any instance of the model by the same algorithm, and being able to scale to rather large instances. This comes not without price, requiring somehow obtaining the aforementioned symbolic model. Pure RL systems, on the other hand, do not require to have a symbolic model, but lack the advantages of AI Planning, being extremely data hungry, domain specific, requiring adaptation and often retraining from scratch when moving to a sufficiently different task. Further, RL algorithms often struggle with problems with large state spaces and sparse rewards. This project aims at marrying between symbolic and reinforcement learning based sequential decision making, studying both the theoretical aspects as well as the practical application of the theory.
