Project-Based Learning in a Deep Learning Class

In this blog, I want to highlight an educational practice that made one of the toughest courses at NYU Tandon interesting and challenging for me: Project-based learning.

In my second semester, I took a Deep Learning class with Prof. Chinamy Hegde. He made the heavy coursework manageable by organizing the class into three projects.  Since this course had 400 students, these projects weren’t regular assignments. Instead, they were competitions hosted on Kaggle. The objective was to build and optimize deep learning models to achieve the best accuracy. The professor would release 50% of the test dataset, which would help rank the code's accuracy, to juggle the leaderboard. 

Snapshot of Kaggle Leaderboard

As these projects required fine-tuning and optimizing models, it was quite a challenge for every student. The leaderboard was determined based on a partially-released dataset. Students were allowed multiple submissions to select the best code in their opinion. Due to the competitive nature, I felt a rush to reach the top of the leaderboard.  The constant shifts in the leaderboard gave a push and at the same time developed curiosity about what techniques other students had used to beat them. The incentive-based grading for the top three teams pushed every student to maintain or improve their rank, and I feel this strategy really helped me stay motivated and continuously learn something new. At the end of two weeks of the project, the remaining 50% of the dataset would be released and the entire leaderboard consisting of more than 120 teams would have new ranks based on the complete test dataset.

Through this process, I connected more with my peers, got feedback on my problem-solving approach, and exchanged ideas on tackling the same challenge in different ways. What excited me most was being pushed to write the best possible solution under time constraints, making the course feel like a semester-long hackathon. I ended up researching and experimenting far beyond what was taught in class, something I wouldn’t have done if the assessments were limited to quizzes or exams. This experience showed me the real strength of project-based learning. It didn’t just test what I knew, but it inspired me to explore what I didn’t.