Monday, March 28, 2016

Episode 52: Scientific Python


We are sponsored by audible!

We are on Patreon!

Intro topic: Virtual Reality


Book of the Show

Tool of the Show

Scientific Python

  • Interfaces
    • Command-Line
    • Spyder
    • Jupyter (formerly ipython notebook)
      • Divides programs into cells and freezes the state of the system in between cells
      • Similar to mathematica notebook
  • Tensor libraries
    • Numpy
      • MATLAB-like interface in python
      • Built on top of LAPACK and other C & Fortran libraries
      • Faster than native python
    • Theano
      • Designed to use the GPU
      • Deferred Execution (batching)
      • Built on top of CUDA (GPU) or Numpy (CPU)
    • Tensorflow
      • Can use either the CPU or GPU
      • Visualization tools (control flow diagrams)
      • Better documentation than Theano
  • Scientific Libraries
    • SciPy
      • Integration, optimization, signal processing, statistics
      • Sparse linear algebra
    • Pandas
      • Data frames, reshaping and pivoting
      • Reading/writing CSV, SQL, HDF5
    • PyMC
      • Bayesian statistical models
      • Markov-Chain Monte-Carlo
  • Visualization
    • Matplotlib
      • Lightweight, can embed graphs in jupyter
    • Panoramix
      • A full service with lots of features, designed to deeply analyze datasets