Notes from reading NNVM code, algorithm analysis, and reflections.
Introduction to CUDA persistent kernel programming paradigm.
Notes of commonly used linker options and what they mean.
Analyze the C++ template tricks embodied in the MShadow matrix operation library.
A brief analysis of how a static computation graph system like MXNet implements LSTM modeling and training.
From here begins the journey of mercilessly thrashing the compiler with C++ templates...
Summary of formula derivations in common deep learning networks.