Selectively Planning with a Flawed Predictive Model

One reasonable way for an artificial agent to learn to make good decisions is for it to learn a predictive model of its environment so it can “imagine” possible futures and plan its behavior to obtain positive outcomes. However, in the inevitable event that the model has flaws, and therefore makes incorrect predictions, the agent may make catastrophically bad decisions. In recent work, Prof. Talvitie and collaborators have been exploring how one might detect errors in the model and selectively use the model to take advantage of what it can predict accurately, while avoiding catastrophic planning failure from what it cannot. The plan for this summer is to continue building on some promising results, potentially including extending these ideas into more complicated model architectures or planning methods, using uncertainty measures to mediate between multiple models with different biases and limitations, or exploring ideas that we haven’t had yet!

(Note: if you are a US Citizen, you are encouraged to apply through https://etap.nsf.gov/award/1783/opportunity/10312; otherwise keep going through URO. Applications from both sites will be considered).

Name of research group, project, or lab
L.A.C.E. Lab
Why join this research group or lab?

Learning Agents in Complex Environments (L.A.C.E.) Lab seeks to understand how artificial agents can be designed to flexibly learn to behave competently in a wide variety of complex, high-dimensional environments. 

One reason to study this is that programs with this capability would be super useful! Flexibility and robust learning are key to many of the short- and long-term ambitious of artificial intelligence. If we ever want artificial agents that clean our houses, serve as personal assistants, and otherwise make sound decisions in our complicated and ever-changing world, it seems like they will have to be able to learn and reason about the world.

Another reason is that it is just so fundamentally weird that humans and other animals can just walk around in a world as complicated as this one and somehow get things done! By studying the problem of creating computational artifacts that can learn and behave flexibly, we may learn about the fundamental computational challenges that need to be overcome in order to make life in a complicated world possible.

Our work will of course not reach these grand ambitions in the short term, but every journey starts with one step....

Representative publication
Logistics Information:
Project categories
Computer Science
Artificial Intelligence
Machine Learning
Student ranks applicable
First-year
Sophomore
Junior
Senior
Student qualifications

To make the most out of our collaboration, student researchers should

  • have experience with and interest in the fundamentals of supervised learning and/or reinforcement learning, most commonly demonstrated by success in a course that covers at least one of these subjects, though experience outside of formal coursework is also valued, and
  • have experience or strong interest in pursuing open-ended problems that require creativity, tenacity, and careful, systematic problem-solving. 

Though not strictly required, being a comfortable, confident C++ programmer is helpful for contributing to this project. 

Hint: In general, if you are interested but aren't sure whether you have the qualifications, just go ahead and apply!

Time commitment
Summer - Full Time
Compensation
Paid Research
Number of openings
3
Techniques learned

Students in the lab will choose a project within the broader research agenda that aligns with their interests and goals and will get lots of practice with and mentorship in many of the stages of basic research in AI/ML, including posing questions, studying related literature, designing algorithms and experiments for scientific insight, interpreting results, and communicating findings. Students will also gain understanding of and experience with algorithms and methodologies for supervised learning and reinforcement learning.

Project start
May 26 (negotiable)
Contact Information:
Mentor
etalvitie@hmc.edu
Associate Professor of Computer Science
Name of project director or principal investigator
Erin J. Talvitie
Email address of project director or principal investigator
erin@cs.hmc.edu
3 sp. | 1 appl.
Hours per week
Summer - Full Time
Project categories
Machine Learning (+2)
Computer ScienceArtificial IntelligenceMachine Learning