This is a simple page to provide basic information about the 2016 Doctoral Consortium at AAAI 2016.
Last update at 11-27-2018 @ 10:47 PM (EST), information current as of then.
This program is preliminary and is likely to be updated several times before the DC. Please check back to ensure you have the latest version.
Location: 211A/B on the 2nd level of the Phoenix Convention Center. This is the West Building, the section closest to the Hyatt Regency Phoenix, on Second Street
The names shown are in order of Student Participant (Mentor).
Friday 12 February 2016
09:00 - 09:40: Introduction
09:40 - 10:20: "Affective Computing of Image Emotion Perceptions" Sicheng Zhao, Harbin Institute of Technology (Dave Roberts, North Carolina State U.)
10:40 - 11:20: "Apprenticeship Scheduling for Human-Robot Teams" Matthew Gombolay, MIT (Mark Roberts, NRC/NRL)
11:20 - 12:00: "Analogical Generalization of Linguistic Constructions" Clifton McFate, Northwestern U. (Laura Hiatt, Naval Research Laboratory)
12:00 - 12:40: Lunch (For DC students and mentors only)
12:40 - 13:20: "Writing stories with help from Recurrent Neural Networks" Melissa Roemmele, U. Southern California (Mark Riedl, Georgia Tech.)
13:20 - 14:00: "Multi-Modal Learning over User-Contributed Content from Cross-Domain Social Media" Wen-Yu Lee, National Taiwan U. (Jingrui He, Arizona State U.)
14:00 - 14:20: Break
14:20 - 15:00: "Machine Learning Methods for Computational Psychology" Sarah Brown, Northeastern U. (Sven Koenig, U. Southern California)
15:00 - 15:40: "Robust Classification under Covariate Shift with Application to Active Learning" Anqi Liu, U. Illinois, Chicago (Kiri Wagstaff, JPL/CalTech)
15:40 - 16:00: Break
16:00 - 16:40: "Privacy Management in Agent-Based Social Networks" Nadin Kokciyan, Bogazici U. (Francesca Rossi, U. Padua)
16:40 - 17:00: Free Time
17:00 - 18:00: Careers Panel (open to all AAAI student attendees)
Panelists: Henry Kautz, Francesca Rossi, Laura Hiatt, and Ugur Kuter
18:00-20:00: AAAI Student Activity in the Atrium Lobby and Arcade (19:00 DC students depart for dinner)
19:30: DC Dinner:Location: bliss/rebar
901 North 4th Street
Phoenix, Arizona 85004
Reservation made under "David Roberts"
The restaurant is 0.7 miles (~15min walk) from the convention center. We will walk over there as a group.
Saturday 13 February 2016
09:00 - 09:40: "Robust Learning from Demonstration Techniques and Tools" William Curran, Oregon State U. (George Konidaris, Duke U.)
09:40 - 10:20: "Estimating Text Intelligibility via Information Packaging Analysis" Junyi Jessy Li, U. Pennsylvania (Mausam, IIT Delhi & U. Washington)
10:40 - 11:20: "Integrating Planning and Recognition to Close the Interaction Loop" Richard Freedman, U. Massachusetts, Amherst (Henry Kautz, U. Rochester)
11:20 - 12:00: "Adapting Plans through Communication with Unknown Teammates" Trevor Sarratt, U. California, Santa Cruz (Milind Tambe, U. Southern California)
12:00 - 13:00: Lunch (For DC participants only)
13:00 - 13:40: "Interactive Learning and Analogical Chaining for Moral and Commonsense Reasoning" Joseph Blass, Northwestern U. (Matthias Scheutz, Tufts U.)
13:40 - 14:20: "Scaling-up MAP and Marginal MAP Inference in Markov Logic" Somdeb Sarkhel, University of Texas, Dallas (Stefano Ermon, Stanford U.)
14:20 - 14:40: Break
14:40 - 15:20: "Architectural Mechanisms for Situated Natural Language Understanding in Uncertain and Open Worlds" Tom Williams, Tufts U. (Cynthia Matuszek, UMBC)
15:20 - 16:00: "Unsupervised Learning of HTNs in Complex Adversarial Domains" Michael Leece, U. California, Santa Cruz (Ugur Kuter, SIFT)
16:00 - 16:20: Break
16:20 - 17:00: "Pragmatics Aware Querying in Heterogeneous Knowledge Graphs" Amar Viswanathan, Rensselaer Polytechnic Institute (Mentor: Héctor Muñoz-Avila, Lehigh University)
17:00 - 18:00: Invited Talk: Héctor Muñoz-Avila (Lehigh University & NSF)
18:00: Close of DC
Sunday 14 February 2016
16:00-18:00: Poster setup (West 301B/C, 3rd Level of Phoenix Convention Center)
18:30-20:30: AAAI-16 Poster Session (DC, EAAI, AAAI-16 technical posters and demos)
We are most grateful to assistance from the Program Committee for reviewing submissions.
- Noa Agmon, Bar-Ilan University
- Bo An, Nanyang Technological University
- Philip Chan, Florida Institute of Technology
- Sonia Chernova, Worcester Polytechnic Institute
- Hana Chockler, King's College London
- Brad Clement, NASA JPL
- Jesse Davis, Stellenbosch University
- Jana Doppa, Washington State University
- Eric Eaton, University of Pennsylvania
- Edith Elkind, University of Oxford
- Piotr Faliszewski, AGH University of Science and Technology
- Jeremy Frank, NASA
- Maria Gini, University of Minnesota
- Benjamin Hirsch, Khalifa University
- Ayanna Howard, Georgia Institute of Technology
- Henry Kautz, University of Rochester
- Chris Kikenveldt, University of Texas at El Paso
- Sven Koenig, University of Southern California
- Maxim Likhachev, Carnegie Mellon University
- Daniele Magazzeni, King's College London
- Tim Miller, University of Melbourne
- Andrea Omicini, Universita Di Bologna
- Jeffrey Rosenschein, Hebrew University Jerusalem
- Jonathan Rowe, North Carolina State University
- Jivko Sinapov, University of Texas at Austin
- Bill Smart, Oregon State University
- Brian Smith, Drexel University
- Adam Smith, Microsoft
- Sebastian Stein, University of Southampton
- Nathan Sturtevant, University of Denver
- Kiri Wagstaff, NASA JPL
- Toby Walsh, National ICT Australia and University of New South Wales
- William Yeoh, New Mexico State University
- Marie desJardins, University of Maryland, Baltimore County
- Each participating student will be assigned a time to give a talk (see the preliminary schedule above). Your talk should be about 25 minutes long. Afterwards, there will be 15 minutes allocated for questions and discussion. The room will be equipped with a data projector, so plan to bring slides (PDF format recommended) on a USB flash drive or a laptop.
- Note that 25 minutes is not a long time. If you have not given many talks before, then we recommend that you practice your talk before coming to the conference.
- Please keep your audience in mind: while we all have a background in AI, there are participants from a wide variety of subareas, so setting the context is critical. Be sure to cover background and motivation before getting into the technical details of your approach. Make sure that the high-level issues in your research are understandable to the audience in general. Having a running example is often a good idea.
- Here is a list of the kinds of questions/comments that past students have received after their talks, which you might want to try and address in your presentation:
- What is your primary contribution?
- How do you know when you'll be done with your thesis? What are the "exit criteria?"
- How do all of the pieces of your thesis fit together?
- Do you have a simple example showing how it works?
- Have you looked at work in subfield X that addresses similar problems?
- Wasn't this already done by person Y?
- How do you expect to address problem Z in your future work?
- How will you evaluate the approach?
- Is your evaluation plan sufficient? Why not look at other/more problem domains?
- Do you expect similar results for other problem domains?
- Is there a real world problem that could benefit from using your work?
- Here is a possible outline for a presentation:
- Problem statement (what problem are you addressing?)
- Motivation (why should we care about this?)
- Your contribution and approach
- Related work (Have others addressed the problem, or how is your problem related to those addressed by others? How is your approach different, and why do you think it might be better?)
- Results so far
- Future work (what else do you expect to do in order to complete your thesis?)
- Summary, Conclusion
- Acknowledgments (list your advisor, any funding sources, and any peers you work with on your project)
- Backup material (see below: material/details that you didn't have time to include but could be useful to help answer questions)
- Below are some more general recommendations for preparing talks:
- If giving talks is a completely new experience for you, then write down an entire script for the talk, broken into sections for each slide. You won't read this script at the talk, but the exercise will help you figure out what you want to say on each slide. If you read the script through using a slow and clear voice, then you can time your delivery and see how well you fall within the prescribed length. People usually get nervous during talks, especially if they are new at it, and that usually means that they talk faster. So keep that in mind when completing this exercise.
- If you are not brand new at giving talks, then create a short outline and write down 1-2 phrases for each slide indicating the main thing(s) you want to convey on each slide. This outline should fit on one page, so you could print it out and have it with you when you present, as a quick reference. It's also helpful to have in case you are nervous and/or distracted during your presentation, because it serves as a quick reminder of what you wanted to say on each slide.
- Read about How to Give a Bad Talk, e.g., Here, here, or here!
- Proofread your slides for typos!!!
- Put page numbers on your slides. This will be especially helpful for the mentor who is leading the discussion, because they can easily refer back to particular slides.
- A good rule of thumb is to have one slide per 2-3 minutes allocated for a talk, e.g., about 8-12 slides in this case. This includes a title page at the beginning and an acknowledgments page at the end.
- It is good to have backup slides that may help you quickly answer questions and allow more time to get feedback. For example, if you have detailed mathematical formulae or software architecture drawings, put these *after* your acknowledgments slide and have them ready to support your answers to questions at a detailed level. Often, having too much technical detail in a presentation can detract from the main point you are trying to get across, so it is better not to include them in the main flow of your talk.
- Beware of cascading bullets in MS Powerpoint (i.e., which require you to hit "next" multiple times on the same slide). This is a matter of stylistic preference, but be aware of the drawbacks (below) if you decide to use this "feature" of your presentation software package. First, as a presenter, you will be distracted from what you are saying by the continual need to interact with the laptop (even via remote control) in order to activate the next cascading bullet. Second, as audience members, we will have to keep returning our attention to the projected slides, away from you and away from what you are saying, in order to read the new bullet. Third, as a presenter, if you need to speed up or go back to a slide during the question/discussion period, it is much easier to do this without the cascading bullets.
- Finally, DON'T STRESS! You are not supposed to have all the answers. If we thought you did, then your application probably would have been rejected. Sometimes "I don't know" is the right answer!
Note: this information is subject to change.
The AAAI-16 Poster/Demo Session will include Doctoral Consortium posters, Technical Paper posters, and demos on Sunday, February 14 from 18:30 - 20:30.
Here are the guidelines for the session:
- AAAI will provide fabric-covered boards and mounting supplies (push pins and velcro). AAAI will provide fabric-covered boards and mounting supplies (push pins and velcro). The poster dimensions for each presenter will be 4 feet high by 4 feet wide. Because there will be two posters side-by-side on boards that are 8 feet wide, we suggest orienting your poster in portrait mode, for instance with dimensions 3.5' high by 3' wide. But you may use the 4x4 space however you choose.
- You are responsible for mounting your material on the board for presentation and may choose any open space in the Doctoral Consortium section. This task should be completed NO LATER THAN 6:00 PM on Sunday evening.
- A good poster allows someone to grasp quickly what your research is all about, and allows you to explain your ideas to them in more detail in case they are interested.
- A representative from the author team should remain by the poster board during the entire session to answer questions and clarify statements. The session should be used as an opportunity to probe deeper into your work, and give attendees an opportunity to ask questions, so your presence is essential. If you would like to keep your poster material, please be sure to retrieve it promptly at 10:30 PM on Sunday night at the conclusion of the poster session. Otherwise, it will be discarded.
Please do not hesitate to contact us if you have any questions about your poster presentation.
Here are some suggestions from DC students in previous years:
- "Start early and practice your presentation multiple times with a group of critical peers and faculty. My presentation was far better thanks to multiple iterations that incorporated feedback from (helpfully) critical lab mates."
- "Focus more on motivation and broad impact of your work. Provide a clear contribution of your work and discuss how it will push the state of the art in AI."
- "Be proactive in interacting with people you don't already know."
- "Embrace the challenge of marketing your work to an unspecialized audience - discuss intuitions, not equations!"
- "Keep the slides simple, talk more about background and intuitions."