Performance Python: Seven Strategies for Optimizing Your Numerical Code
summary
Seven concrete tactics for speeding numerical Python with a K‑means case study line profiling shows label computation takes 93% of time and vectorizing with NumPy vectorization flips hotspots so runtime goes 7.4s down to 0.13s; specialized structures (KD‑tree, Pandas groupby) shave another 30% Cython and Numba compile hotspots to milliseconds (Cython ~100ms → ~2ms Numba ~100ms → ~1.5ms) Dask enables out‑of‑core and distributed runs but adds overhead and scikit‑learn gives a robust ready‑made K‑means.
AI-written summary · about
chapters (4)
related talks
34mNumPy: vectorize your brain
38mVectorize all the things! Using linear algebra and NumPy to make your Python code lightning fast
35mVectorization: How slow Python runs fast code
22mExperiences making CPU-bound tasks run much faster
01h 33mPower up your work with compiling and profiling
40mSpeed Without Drag
01h 12mFurious & Fast Python 7: Writing Fast Python Code
29mVectorize using linear algebra and NumPy to make your Python code fast