4 months ago
Second semester project for the Interactive Machine Learning 2021/2022 course
This is the second semester project for the Interactive Machine Learning 2021/2022 course. The task is to choose optimal first batch of queries for training a simple classification model.
The goal of this competition is to choose three subsets of samples from the data pool such that they compose the best possible initial data batch for active learning of a simple prediction model (logistic regression).
Competition rules are given in Terms and Conditions.
The description of the task, data, and evaluation metric is in the Task description section.
Participants of the challenge are obliged to follow the competition rules:
- This challenge is organized by Andrzej Janusz and Daniel Kałuża (the Organizers) for students enrolled in the Interactive Machine Learning 2021/2022 course at the Faculty of Mathematics, Informatics, and Mechanics at the University of Warsaw.
- The provided data sets are the property of the Organizers and the KnowledgePit platform. It is forbidden to share or redistribute provided data sets to any third party without explicit consent from the Organizers.
- Each team in the competition may consist of only one person. Working in larger groups or sharing solutions with other teams is strictly forbidden.
- Each team has a limited number of submissions - the limit is set to 100.
- The number of submissions per day is limited to 10.
- Participants can use data that was made available in the challenge - using any external resources is forbidden. Queries regarding the external resources need to be issued through the competition forum.
- It is strictly forbidden to hack the provided data or to exploit any unfair data leak that can improve the solution score. All attempts at making predictions for any test instance using information extracted from other sources will result in disqualification.
- The deadline for submitting the solutions is May 29, 2022 (23:59 GMT). Late submissions will not be accepted.
- Each team is obliged to provide a short report describing their final solution. The report must contain information such as the name of the team, the names of all team members, and a brief overview of the used approach. It should be submitted using the KnowledgePit submission system by May 29, 2022 (23:59 GMT).
- By enrolling in this competition, you grant the Organizers the right to process your submissions and reports for the purpose of evaluation and post-competition research.
- The final project score will depend on the quality of the solution (the score obtained in the final evaluation), and on the quality of the submitted report.
|Rank||Team Name||Is Report||Preliminary Score||Final Score||Submissions|
The task in this project is to choose three subsets of samples from the data pool such that they allow constructing the most accurate logistic regression model. The size of subsets should be 100, 200, 500, respectively.
The data pool is given as a single csv file. In particular, each line of the file corresponds to a description (a feature vector) characterizing approximately 1.8 seconds of readings of inertial sensors placed on seven different body parts of firefighters during exercises. In total, the data pool has 20000 samples described by 1680 features. The labels for this data are different classes of actions that were conducted by firefighters.
Each sample can be assigned to one of 16 classes: "ladder_going_down" , "ladder_going_up", "manipulating", "no_action", "nozzle_usage", "running", "searching", "signal_hose_pullback", "signal_water_first", "signal_water_main", "signal_water_stop", "stairs_going_down", "stairs_going_up", "striking", "throwing_hose", "walking".
No labels for the training data are available.
Format of submissions: solutions should be submitted as text files with three lines. The first line should contain exactly 100 integers - the indices of samples from the data pool (samples are indexed starting from 1), separated by commas. The second and the third line should contain analogous indices for the second and the third set of samples, with sizes 200, and 500, respectively.
Evaluation: the evaluation of submitted solutions will be done using a LASSO logistic regression model, trained independently on the three sets of samples added to the initial data which was made batch available to participants. Each model will be evaluated on a separate test set (hidden from participants). The quality metric used for the evaluation will be the average BAC. From each score, we will compute balanced accuracies of models trained only on the selected initial data batches and the three results will be averaged with weights 5, 2.5, and 1 for the subset sizes 100, 200, and 500, respectively.
During the challenge, your solutions will be evaluated on a small fraction of the test set (10%), and your best preliminary score will be displayed on the public Leaderboard. After the end of the competition, the selected solutions will be evaluated on the remaining part of the test data and this result will be used for the evaluation of the project.
The LASSO logistic regression model will be trained with the regularization parameter lambda set to 0.01.