University of Michigan - Ann Arbor

NumPy is a Python library for scientific computing that contains support for some fundamental data structures used by scikit-learn, such as multi-dimensional arrays. Example import: `import numpy as np`

NumPy

In NumPy, there are two objects that are crucial to understanding how this library works. Those are:

1. N-dimensional array object (ndarray)

2. Universal function object (ufunc)

However, keep in mind that there are other objects that are built upon these that can be incorporated into whatever you're doing.

Fundamental Objects in NumPy

Developers need to know what works and how to use it. This site uses less math and more tutorials with working code.
https://machinelearningmastery.com/

Machine Learning Mastery

NumPy Documentation

Typically, data that's input to scikit-learn will be in the form of a NumPy array.

NumPy vs. Scikit-learn

What does the following code do?

Recall that "np.dot(a,b)" performs a matrix multiplication on a and b, whereas "a*b" performs an element-wise multiplication. Consider the two following random arrays "a" and "b": what is the shape of c?

Consider the following code: what will be c?

Consider the two following random arrays "a" and "b" below: what will be the shape of "c"?

Because the Numpy array is densely packed in memory due to its homogeneous type, Numpy gets around 30 times faster than Python.

Make big difference when handling big data.

Numpy Speed


In python, there  is already data types that it has by default such as boolean, float, integer, etc. However, NumPy offers additional data types:

- i - integer
- b - boolean
- u - unsigned integer
- f - float
- c - complex float
- m - time delta
- M - date time
- O - object
- S - string
- U - unicode String
- V - fixed chunk of memory for other type (void)

Data Types in NumPy

Within the NumPy library, there are several types of mathematical functions built-in. These functions are typically used across elements of a NumPy array.

Types of Mathematical Functions:
- Trigonometric
- Rounding
- Exponents and Logarithms
- Arithmetic
- Other Special Functions

Mathematical Functions in NumPy

Unlike the arrays supported by NumPy, the columns in a DataFrame can be of all different types. You can have one column holding string values, another column holding a date, another column holding a floating point number and so on.

Pandas Dataframe vs. numpy.array

In an n-dimensional array there are two characteristics that define it:
- Shape of the array
- type of element the array is composed of

Essentially, ndarrays describes the collection of items of the same  data type. Since that's the case, then every item will take up the same size block of memory. Each of those blocks are interpreted in the same way.

Ndarray Object

Essentially, there are NumPy functions that operate on ndarrays. All (ufuncs) wrap a core function that takes scalar inputs and produces scalar outputs.

Why would this be useful?
- Say for example a user wants to sum individual elements in a list with the same corresponding elements in another list. One could iterate through both lists and add them as it goes through the iteration. Or, if the user uses a ufunc called "add(x,y)"  then it does the same thing - but saves times.

Learn Before

Related