Performing out-of-core computations on large arrays with Dask

Dask is a parallel computing library that offers not only a general framework for distributing complex computations on many nodes, but also a set of convenient high-level APIs to deal with out-of-core computations on large arrays. Dask provides data structures resembling NumPy arrays (dask.array) and Pandas DataFrames (dask.dataframe) that efficiently scale to huge datasets. The core idea of Dask is to split a large array into smaller arrays (chunks).

In this recipe, we illustrate the basic principles of dask.array.

Getting ready

Dask should already be installed in Anaconda, but you can always install it manually with conda install dask. You also need memory_profiler, which you can install with conda install memory_profiler.

How to do it...

1. Let's import the libraries:

   >>> import numpy as np
       import dask.array as da
       import memory_profiler
   >>> %load_ext memory_profiler

2. We initialize a large 10,000 x 10,000 array...