Book: Efficient Data Science in Python

Participant(s)

Course Details

Profiling and timing code: The process of identifying bottlenecks in code is the first and most important step to improving efficiency. Here we will see how we can identify our bottlenecks, and quantify future improvements.
Vectorising computations: Here we will discuss why vectorisation is efficient, and see how we can vectorise code with NumPy and pandas. We also see how we can use BLAS libraries to speed up our code for free.
Just-in-time compilation: JIT compilation has the potential to drastically improve the performance of your code, whilst requiring minimal alteration. Here we review the PyPy and Numba project’s JIT support.
Cython: A discussion of compiled vs. interpreted languages, and their associated performance. An introduction to Cython and its static type declarations.

By the end of the course participants will be able to…

identify bottlenecks in the performance of their code
replace costly for-loops with vectorised NumPy and pandas operations
select a BLAS library appropriate for their system
use PyPy and Numba to bring just-in-time compilation to their data-science workflow
understand Cython’s static type declarations and be able to write and execute Cython code.