This is a PAN BS event. All students enrolled in the IITM Data Science or Electronic Systems programme are welcome to register for our event.
Absolutely! We strongly encourage you to join us. No need to worry if you're just starting out and don't have a coding background. We've got you covered with ML-101 (tutorial/mentor) sessions and beginner-friendly resources to help you get started. The goal isn't to master ML or become an expert overnight but to gain a solid understanding of the basics and explore how ML works. This will be a fantastic hands-on opportunity and learning experience. So, take the plunge. You'll be the happiest when you build your first model :)
Yes. Margazhi 2025 will be conducted virtually. Registration for LogicLooM 3.0 is mandatory for attending the talk sessions.
Maximum 3.
Collaboration is an important aspect of ML contents and working as a team offers numerous benefits - not only does it save time but also help you develop essential peer-connect and teamwork skills. Plus, events like these are about learning and sharing, not competing individually like in a graded exam. The true essence lies in collaborating, exchanging ideas and growing together. So we highly recommend working together... it'll make the experience more enriching and enjoyable!
That's no problem at all! If you already have a teammate in mind, simply ask them to register individually. Team formation will take place through our dedicated event portal where participants can invite others to join their team and the invitees can accept to become a part of it. It's simple and streamlined - just register and the rest will follow!
Yes. While teamwork is encouraged, we understand that some participants may prefer to compete individually. If you have the necessary skills like proficiency in Python, machine learning and NLP, feel free to go solo.
Simply select the option 'I would like to be assigned to a team by the organizers' on the registration form. In such cases, our system will match you with others based on the rules, prioritizing participants of the same level.
NO! That's strictly not allowed.
We will be using ROUGE-L and BLEU metrics. Event specific details will be shared with the registered participants in due time.
Not really. We have carefully reviewed the nature of the task and dataset size to ensure that all code can be executed seamlessly on the free version of Google Colab. You are free to make use of Pro+ accounts if you have. However account subscriptions loaded with GPU credits will not be provided by the organizers.
No restrictions. Teams are allowed to use pre-trained models, open-source code for building their solutions.
Cross-team collaboration is NOT allowed. Participants in a team are supposed to rely on their own skill and knowledge, with no help from other teams or anyone else.
Yes! Usage of external data is NOT allowed.
Yes, mentoring sessions will start after the end-sem itself. Additionally, assistance sessions will be conducted to help the teams (especially the freshers) during the model training and development phase.
To access the participants portal, a seamless login system is enabled through Google OAuth integration. Simply sign in using the IITM student G-account used while registering for our event on the Paradox portal. If you attempt to log in with a different account, you'll receive an error message and won't be able to access it. Please write to logicloom@ds.study.iitm.ac.in for any queries.
We suggest you to read the event rulebook first, everything has been explained in detail. Still, if you have any general questions or face technical issues, feel free to contact the organizers at logicloom@ds.study.iitm.ac.in. We will try to reply back within one working day.
Exciting cash rewards and goodies worth 20.5K await! 💰 👀 :)
ML 101 Sessions - 29th Dec 2024, 2nd and 4th Jan 2025
Participants, especially Begineer Pioneers are strongly encouraged to attend the mentoring sessions. These sessions will provide in-depth guidance, explaining key concepts from scratch and offering one-on-one support during the model development phase. Join us at meet.google.com/xwr-wqyc-ssj
Release of Train data - 2nd Jan 2025
A 3k row labeled training dataset will be released along with the detailed problem statement, milestones and information on evaluation metrics. Teams can start exploring the dataset, build & train suitable models. External datasets CANNOT be used in the competition.
Release of Labeled development set - 2nd Jan 2025
A development set containing 1k new samples will be provided for model evaluation. Participants can use this set to test their model-generated values against the actual data, using metrics like ROUGE-L and BLEU scores to assess performance and select the best model or further optimise and fine-tune their model's hyperparameters.
Release of Unlabeled Test set & Submission of Deliverables - 4th to 5th Jan 2025 10.00 IST
Teams will use their best-performing model to generate predictions for an unlabeled (ground truth values are not revealed) test set. Each team is allowed to submit one run, which should include the predictions (in .csv format) generated by their model and working code (in .ipynb format) that can be used to reproduce their results.
App-deployment & Presentation - 6th Jan 2025
Here's a twist. There will be another surprise test set which will features content that is significantly different in style, genre, or domain of the other sets - this challenges participants to handle more generalization, adaptability and robustness. The top-performing teams will proceed to a final round where they will deploy their model and deliver a presentation. Evaluation will consider model performance, novelty, techniques used, and overall presentation quality.