...
Task: For each of the remaining 2800 runs, predict where the bot is on the 100th step. The competition will be hosted on in-class-Kaggle.
Kaggle Link: Coming soon... https://www.kaggle.com/t/a159a375ad704dba8a233abd2340f729
Download the data below as zip file. When unzipped you will find the two files, obervations.csv and labels.csv
...
Due date The deadline is 11:59 pm, SaturdayMonday, 3rd 5th December. The due date for the report on CMS will be announced soon and is a couple days after the competition closes on Kaggle. Submit what you have at least once by an hour before that deadline, even if you haven’t quite added all the finishing touches — CMS allows resubmissions up to, but not after, the deadline. If there is an emergency such that you need an extension, contact the professors.
...
- Report: In the end of the competition each group should submit a 5-15 page writeup that includes visualization, clear explanation of methods etc. See grading guidelines for details about what is expected from the writeup. (worth 50% of the competition grade)
- Predictions: Competition is held on Kaggle in-class as a competition. You can submit your predictions to kaggle to compete with your friends. You should also submit your predictions on CMS. (worth 40% 50% of the competition grade)
- Code: Submit the code you used for kaggle as a zip file.
...
2. Footnote: We make an exception for sources that can be taken for granted in the instructional setting, namely, the course materials. To minimize documentation effort, we also do not expect you to credit the course staff for ideas you get from them, although it’s nice to do so anyway.
Grading Guidelines:
Grading:
- Clear explanation of your main model (20 points)
- Explain any preprocessing you did, explain clearly what your model takes as input
- Explain clearly what algorithm you used to train and not just the model
- How does your model fit the problem description? (10 points)
- How does model account for the fact that there were 3 bots?
...
- (10 points)
- How were parameters chosen in a principled fashion? (10 points)
- Failed attempts. (Have a clear flow of your reasoning for why you tried various models and how their failure guided you to pick next one) Give clear comparison of things you tried. Dont go for numbers but rather clear progression of thought and how each model guided the next. (15 points)
- Visualization (what did you learn from them and how they guided you). This includes tables, plots, graphs etc. (10 points)
- Supervision: How did you use the labeled examples given in your model. Did you use these to minimize kaggle submissions? (10 points)
- Unlabeled examples: How were the unlabeled data points part of you model (10 points)
- Understanding data, what did you learn from the observations and how was it used in your approach? (5 points)
Bonus (at the discretion of the graders):
Tried new or more methods not necessarily covered in class
Developed new algorithm or methods, tweaked existing methods to fit the problem better