Python's Bright Future in Science
summary
Python already displaces IDL/Matlab in papers, powered by NumPy/SciPy for things like three‑line quantile normalization and fast sparse linear algebra that produces spectral coordinates. Raw Python can be 100x–900x slower than C, but Cython and C‑extensions wrap native code and Numba JIT turns hot loops into C‑speed (example: a demo gives ~600x). New work proposes a NumPy intermediate representation and Microsoft’s Pigeon plugin (100% NumPy compatibility on Windows) to let many JITs and GPUs accelerate Python without rewriting libraries.
AI-written summary · about
chapters (27)
- 0:00Introduction
- 1:00Why is Python so popular
- 3:45How to find a pattern
- 4:29Quantile normalization
- 6:38Pipe operator
- 7:13DateDelay
- 8:19Limitation
- 10:50Numpy
- 12:11Sparse matrices
- 13:48Python is fast
- 14:24siphon
- 14:57psyphon
- 16:25integer
- 16:53Julia
- 20:19The Dirty Secret
- 21:57The Future of Python
- 23:53Pigeon
- 25:37The Community
- 26:59Bonus
- 27:13Questions
- 27:50Big Data
- 28:21Optimization
- 31:05Scythe
- 34:40Premature optimization
- 37:05Python vs other languages
- 39:25Type annotations
- 41:00GPU compilation
related talks
45mPython for Speed, Scale, and Science
22mPyPy and the future of the Python ecosystem
48mPython Performance: Past, Present and Future
27mSpeed up Python with hardware accelerators
26mThe Future of Python: Evolution or Succession?
58mPython performance Past, Present, Future
47mBig ideas shaping scientific Python: the quest for performance and usability
43mKEYNOTE: Scaling Out PyData