What the problems asked
Team of three
Linear-algebra modelling under competition conditions: Markov population dynamics with a time-varying-rate correction, the power method and SVD; and the headline task: clean, run PCA on and cluster a set of infrared absorbance spectra, and say how many distinct food samples were really in it.
I entered with no prior linear algebra and built the prerequisite chain from zero in five days.
1,000→~25
d ≈ 8
One thousand spectra resolved to about twenty-five unique samples at intrinsic dimension about eight, using the duality between the covariance and Gram matrices.
Also: co-author, applied mathematics paper on Chennai flood modelling, Feb 2024 (Grade 9).Taught it afterwards: simc-masterclass.vercel.app