3 Minute Thesis Competition 2026
Keywords
Summary
151 words
Critical Evaluation
Value of the Information & Strength of the Argument
The video provides a valuable overview of current research in mathematics, showcasing the diversity and depth of work being done by early-career researchers. Each presentation is well-structured, with a clear statement of the problem, the approach, and the significance of the work. The argumentation is generally solid, with presenters providing justifications for their methods and results. For example, the first presenter explains the need for machine learning to accelerate expensive simulations, and the second presenter demonstrates how a simple mathematical model can isolate the collapse mechanism of an ice cap. The presentations are persuasive and effectively communicate the importance of the research.
111 words
Title / Content Match
The title accurately reflects the content: a 3-minute thesis competition featuring seven presentations.
Quality & Reliability
8/10
The video presents seven research projects by postgraduate students at Oxford, each with clear methodology and results. The content is peer-reviewed in the sense of being presented in a formal academic competition. However, the format (3-minute talks) limits depth, and the claims are not independently verified in the video.
Key Moments
Markers derived by PSI from the transcript: the creator did not define chapters.
- Introduction and first presentation on machine learning for re-entry simulations
- Second presentation on the collapse of the Barnes Ice Cap
- Third presentation on parameter-free accelerated gradient descent
- Fourth presentation on accelerating eigenvalue problems
- Fifth presentation on percolation theory
- Sixth presentation on causal inference with instrumental variables
- Seventh presentation on continuous logic and C*-algebras
- Announcement of winners and conclusion
Cited Sources
- Previous years' competitions playlist — Mentioned in the video description as a resource for previous competitions.
Contribution & Novelties
The video provides a snapshot of cutting-edge research in mathematics, with each presentation offering a novel contribution to its field. The first presentation introduces a machine learning approach to accelerate simulations of rarefied gas dynamics, which is crucial for spacecraft re-entry. The second presents a mathematical model for the collapse of an ice cap, highlighting the concept of irreversibility. The third introduces a parameter-free accelerated gradient descent algorithm with optimal convergence rates. The fourth presents a method for accelerating sequences of eigenvalue problems using subspace recycling and row subsampling. The fifth extends results from Bernoulli percolation to Gaussian percolation, contributing to the understanding of phase transitions. The sixth develops a robust method for causal inference with imperfect instrumental variables. The seventh uses continuous logic to axiomatize subclasses of C*-algebras, providing a framework for quantum mechanics.
Pour aller plus loin :
- Mixture Density Networks — Relevant to the first presentation on machine learning for collision dynamics.
- Navier-Stokes equations — Background for the first presentation on fluid dynamics.
- Barnes Ice Cap — Directly relevant to the second presentation.
- Percolation theory — Relevant to the fifth presentation.
- Causal inference — Relevant to the sixth presentation.
- C*-algebra — Relevant to the seventh presentation.
199 words
Radar Profile
The radar profile shows high scores in quality of information and global reliability, reflecting the academic rigor of the presentations. The quantity of information is moderate due to the short format, and the technical level is high, indicating a specialized audience.