Use the github server to get two starting point directories for this lab. One, with the prefix lab6, will be shared with your partner (if you have one). The other, with the prefix poster, will not be shared.
For this lab, you may work together on an experiment of your choice that relates to the theme of this class. In the lab6 repo, you should add, commit, and push any files related to your experiment and you should summarize how to execute your experiment in the Summary.ipynb.
In the poster repo, you should create your own individual poster (saved as a PDF) based on the experiment.
NOTE: You should not print your poster. We will schedule a time at ITS to do this later, after I have had a chance to review your poster.
Most scientific conferences include poster sessions where researchers present their work. Often these poster sessions can be quite large with many posters on display at once. Conference attendees walk through the poster session, browsing the titles and choosing a few posters to stop at and explore in more depth.
In order to grab people's attention, your poster must be visually engaging. It must have large enough font sizes so that it can read from 10-15 feet away. Most importantly it needs to clearly highlight the key ideas of your work. Typically the poster will be a short description of a longer paper. The goal of the poster is to generate interest in your work, hopefully leading people to follow up by reading your paper.
Here are two example posters from my own research. The first was presented at a small, interdisciplinary workshop as part of a much bigger conference. The audience was largely made up of researchers who were neither computer scientists nor roboticists. The second was presented at a medium-sized conference with a more robot-focused audience, and was based on the paper that we read in the second week of the class.
Poster presented at the "Designing for Curiosity" Workshop as part of the 2017 Conference on Human Computer Interaction in Denver, Colorado
Poster presented at the 2017 Conference on Development and Learning and on Epigenetic Robotics in Lisbon, Portugal
Both of these posters discuss the same research project, but at different phases in the work. In my opinion, one of these posters is both more visually engaging and more effectively highlights the results.
There are many other example posters along the main CS Department hallway, and throughout the entire Science Center. As you walk around the Science Center, notice which posters grab your attention and try to determine why.
During lab today, I will visit with each of you to hear about your proposed research idea. You should have a hypothesis and an experiment planned to test this hypothesis. Hypothesis testing involves comparing a control condition to a treatment condition, where you have varied the experimental set up in some significant way.
Flesh out as many of the details as possible before you begin setting up your experiments. For example, if you are using a supervised neural network, how will you obtain the training data needed? Or, if you are using evolution, what will you use for the fitness function? Or, if you are using reinforcement learning, how will the rewards and punishments be determined? In addition, what are the inputs to the system and will they need to be pre-processed in any way? What are the outputs of the system?
You should do both quantitative and qualitative analyses of your results. For the quantitative results, you will need to run multiple trials, where you keep the parameters fixed, and at the very least do some basic statistical analyses (such as computing the average and standard deviation). If possible, determine whether there are any statistically significant differences in performance. For the qualitative results, you should examine the types of behaviors produced.
Think carefully about how best to visually depict, on your poster, the interesting quantitative and qualitative differences that you discover.
Doug Willen from ITS has developed this helpful video about how to create an effective scientific poster.
Feel free to use whatever software package you prefer, as long as you can generate a PDF of your poster that is the correct dimensions (48 inches wide and 35 inches high). One option is to use Google Slides, which is what I used to create both of the example posters above. If you choose to use Google Slides, here are the steps you'll need to do to get started:
Now you can start adding the content of your poster. These requirements apply to all posters:
You will present your poster in lab next Monday. We will project your poster on the screen at the front of the class. Be prepared to talk for 5 minutes. Give a top-down explanation of your work, starting with the overall goal and then going into the methods used and the results.
Be sure to push all of the files related to your experiment to the lab6 repo and push your individual poster to the poster repo.