My first bug solved in Librosa - waveform display
How a stuck waveform display led to discovering garbage collection issues with Matplotlib callbacks, and why fixing it properly required understanding weak references and memory leaks.
Sometimes, when implementing software, we learn a technology to solve a problem, we ship it as fast as possible and give it little time to be properly digested. Complex topics require time, discussions, note taking and experimentation. I created this page to document some of this process.
How a stuck waveform display led to discovering garbage collection issues with Matplotlib callbacks, and why fixing it properly required understanding weak references and memory leaks.
A discussion on how to speed up repeated linear solves by using a spectral trick. A real-world example, borrowed from physics, of how a simple change in algorithm lead to a significant speedup.
How this concept appears in three distinct fields (Databases, Deep Learning, and Quantum Physics) and got me confused. Here I explore a bit more about it and the fundamental pattern hiding underneath.
No posts match your filters yet. Try a different keyword or category.