2013 picture

Travis Mandel

Assistant Professor
Computer Science and Engineering
University of Hawai'i at Hilo
Email: tmandel [at] hawaii [dot] edu

New! I am very excited to have started as a new professor at the University of Hawai'i at Hilo. I am helping lead the development of the new 'Ike Wai data science program there, including both teaching and research.

Before joining UH Hilo, I earned my PhD from the Paul G. Allen School of Computer Science and Engineering at the University of Washington advised by Zoran Popović and Emma Brunskill (Stanford). Before that, I earned my Bachelor's degree in Computer Science from Carnegie Mellon University.


This fall 2017 semester, I am teaching:
CS150 Intro to Computer Science I
CS440 Artificial Intelligence

In terms of research, I am interested in the broad question of how autonomous agents should best interact with humans to help improve the lives of people across the world. Most of my research focuses on Reinforcement Learning (RL), a machine learning (ML) subfield of Artificial Intelligence (AI) which has enormous potential to help create systems that learn to near-optimally interact with humans. Although there are many interesting domains in this space (such as personalized healthcare), as a real-world RL testbed I have examined the domain of educational games, launching experiments to thousands of students with the goal of using RL to directly improve student engagement and learning outcomes.

However, there are many difficult challenges that we face in applying RL to such human-focused domains, several of which I have examined throughout my PhD studies. First, the problem of offline evaluation: When data is expensive or risky to gather, how can we evaluate different RL approaches in a reliable manner without actually deploying them? Next, the problem of efficient exploration: When we do actually launch such an algorithm, how can we make sure that it learns to improve its performance as quickly as possible, even in the face of real-world issues such as delay? Third, the problem of Human-in-the-Loop RL: How can an RL agent best leverage human creativity and intuition to improve its performance? Developing better solutions to these three problems has the potential to massively improve not just education, but other domains as well.

In the past I have done significant work in channel coding (CRC error correction) for wireless sensor networks (WSNs), a totally unrelated area of computer science.

Selected Publications

Where to Add Actions in Human-in-the-Loop Reinforcement Learning
Travis Mandel, Yun-En Liu, Emma Brunskill, Zoran Popović.
AAAI Conference on Artificial Intelligence (AAAI) 2017.

Efficient Bayesian Clustering for Reinforcement Learning
Travis Mandel, Yun-En Liu, Emma Brunskill, Zoran Popović.
International Joint Conference on Artificial Intelligence (IJCAI) 2016.

Offline Evaluation of Online Reinforcement Learning Algorithms
Travis Mandel, Yun-En Liu, Emma Brunskill, Zoran Popović.
AAAI Conference on Artificial Intelligence (AAAI) 2016.

The Queue Method: Handling Delay, Heuristics, Prior Data, and Evaluation in Bandits
Travis Mandel, Yun-En Liu, Emma Brunskill, Zoran Popović.
AAAI Conference on Artificial Intelligence (AAAI) 2015.

Offline Policy Evaluation Across Representations with Applications to Educational Games
Travis Mandel, Yun-En Liu, Sergey Levine, Emma Brunskill, Zoran Popović.
Autonomous Agents and Multiagent Systems (AAMAS) 2014.

Practical Error Correction for Resource-Constrained Wireless Networks: Unlocking the Full Power of the CRC
Travis Mandel, Jens Mache.
ACM Conference on Embedded Networked Sensor Systems (SenSys) 2013.

Other Publications

Examining PhD Student Interest in Teaching: An Analysis of 19 Years of Historical Data
Travis Mandel, Jens Mache.
Poster at ACM Special Interest Group on Computer Science Education (SIGCSE) 2017.

Crowdsourcing Accurate and Creative Word Problems and Hints
Yvonne Chen, Travis Mandel, Yun-En Liu, Zoran Popović.
AAAI Conference on Human Computation and Crowdsourcing (HCOMP) 2016.

Developing a Short Undergraduate Introduction to Online Machine Learning
Travis Mandel, Jens Mache
Journal of Computing Sciences in Colleges, Volume 32, Number 1. 2016.

Nonstationary Evaluation for Reinforcement Learning
Travis Mandel, Yun-En Liu, Emma Brunskill, Zoran Popović.
Reinforcement Learning and Decision Making (RLDM) 2015.

Trading Off Scientic Knowledge and User Learning with Multi-Armed Bandits
Yun-En Liu, Travis Mandel, Emma Brunskill, Zoran Popović.
Educational Data Mining (EDM) 2014.

Towards Automatic Experimentation of Educational Knowledge
Yun-En Liu, Travis Mandel, Emma Brunskill, Zoran Popović.
Computer-Human Interaction (CHI) 2014.
Honorable Mention

Unbiased Offline Evaluation of Policy Representations for Educational Games
Travis Mandel, Yun-En Liu, Sergey Levine, Emma Brunskill, Zoran Popović.
Data Driven Education Workshop, Neural Information Processing Systems (NIPS) 2013.

Predicting Player Moves in an Educational Game: A Hybrid Approach
Yun-En Liu, Travis Mandel, Eric Butler, Erik Andersen, Eleanor O'Rourke, Emma Brunskill, Zoran Popović.
Educational Data Mining (EDM) 2013.
Best Paper Nomination.

ContextType: Using Hand Posture Information to Improve Mobile Touch Screen Text Entry
Mayank Goel, Alex Jansen, Travis Mandel, Shwetak N. Patel, Jacob O. Wobbrock
Computer-Human Interaction (CHI) 2013.

Investigating CRC Polynomials that Correct Burst Errors
Travis Mandel, Jens Mache
International Conference on Wireless Networks (ICWN) 2009.

Selected CRC Polynomials Can Correct Errors and Thus Reduce Retranmission
Travis Mandel, Jens Mache
Workshop on Information Theory for Sensor Networks (WITS), International Conference on Distributed Computing in Sensor Systems (DCOSS) 2009.

Sensor Network Security: Elliptic Curve Cryptography on SunSPOTs
Jens Mache, Samuel W. Bock, James Elwell, Dennis P. Gosnell, Travis Mandel, Jonathan S. Perry-Houts
International Conference on Wireless Networks (ICWN) 2008.