Roofline Models: From Speculation to Modelling
Performance modelling isn't just guess work. It is a science backed by iterative experimentation and edivence gained.
From Math to Machines
Recorded experiences and insights from the intersection of mathematics and computer science.
Building SVMs from Scratch in C++: Glacier’s PEGASOS Implementation
Implemented PEGASOS SVM classifier and regression models in Glacier.ML - noted 4-10x performance speed-ups in training + inference times against Scikit-learn SGD.
Optimized k Nearest Neighbours: From Naive to Hardware-Aware
Implementated optimizations to kNN Regressor and Classifier models, observed 100x inference speed-ups compared to naive version.
Building my own C++20 Numerical Algorithms library
Introducing Glacier.ML - a C++ based Supervised ML library, written from scratch to understand theoretical ML.