In August 2025, I attended the PiMA 2025 Summer Camp, one of 24 students selected nationwide. With the theme “Mathematics in Generative Models”, we got to know the basic algorithms behind popular generative models, and had the opportunity to learn and present them to everyone. And I also had the opportunity to interact and learn from teachers and seniors with the same passion in a week of deep learning, inspiration and connection.
During our time at the PIMA Summer Camp, we built a strong foundation in linear algebra, probability, statistics, and deep learning, while also developing basic programming skills in Python and learning to write with LaTeX. We then applied this knowledge to explore different aspects of generative models and their real-world applications. The experience culminated in a group project, where, with guidance from mentors at leading universities both within and outside the country, we put our learning into practice in a collaborative and meaningful way.
My teammates and I conducted a project focused on Normalizing Flows, a class of generative models that transform a simple probability distribution into a more complex one through a series of invertible and differentiable mappings. This approach allows for both efficient sampling and exact likelihood evaluation, making it a powerful tool in modern machine learning. In our project, we studied the mathematical foundations of Normalizing Flows, implemented them in Python, and explored their applications in data generation and density estimation under the guidance of our mentors. You can check out our finished report here:
