15.2 NumPy

NumPy is a powerhouse for scientific computing in Python. It provides support for large, multi-dimensional arrays and matrices, enabling fast and efficient operations on datasets. This makes it essential for data science tasks and the foundation for other libraries.

NumPy offers tools for creating and manipulating arrays, from basic indexing to reshaping and joining. It also provides a wide range of mathematical and statistical functions, allowing for efficient data processing and analysis. Understanding these features is crucial for effective data manipulation in Python.

Introduction to NumPy

Purpose and features of NumPy

Top images from around the web for Purpose and features of NumPy

• 

  py numpy库_yichudu-CSDN博客 View original

  Is this image relevant?

• 

  Data science: Reshape and stack multi-dimensional arrays in Python numpy View original

  Is this image relevant?

• 

  Python for Scientific Computing View original

  Is this image relevant?

• 

  py numpy库_yichudu-CSDN博客 View original

  Is this image relevant?

• 

  Data science: Reshape and stack multi-dimensional arrays in Python numpy View original

  Is this image relevant?

1 of 3

Top images from around the web for Purpose and features of NumPy

• 

  py numpy库_yichudu-CSDN博客 View original

  Is this image relevant?

• 

  Data science: Reshape and stack multi-dimensional arrays in Python numpy View original

  Is this image relevant?

• 

  Python for Scientific Computing View original

  Is this image relevant?

• 

  py numpy库_yichudu-CSDN博客 View original

  Is this image relevant?

• 

  Data science: Reshape and stack multi-dimensional arrays in Python numpy View original

  Is this image relevant?

1 of 3

• Fundamental library for scientific computing in Python
  • Provides support for large, multi-dimensional arrays and matrices
  • Offers a wide range of mathematical functions to efficiently operate on these arrays
• Key features enable fast and efficient operations on large datasets
  • Supports broadcasting allows arrays with different shapes to work together
  • Provides tools for integrating C/C++ and Fortran code
  • Enables the use of sophisticated mathematical and statistical functions (

    [np.sin()](https://www.fiveableKeyTerm:np.sin())

    ,

    [np.exp()](https://www.fiveableKeyTerm:np.exp())

    )
• Essential for data science tasks
  • Allows for efficient storage and manipulation of data (

    np.array()

    ,

    [np.zeros()](https://www.fiveableKeyTerm:np.zeros())

    )
  • Serves as the foundation for other data science libraries (Pandas, SciPy)

Creation and manipulation of arrays

• Create arrays using

  np.array()

  from a list or tuple
  • Create arrays with specific values using functions (

    np.zeros()

    ,

    [np.ones()](https://www.fiveableKeyTerm:np.ones())

    ,

    [np.arange()](https://www.fiveableKeyTerm:np.arange())

    )
  • Specify the data type of an array using the

    [dtype](https://www.fiveableKeyTerm:dtype)

    parameter (

    np.array([1, 2, 3], dtype=np.float64)

    )
• Array attributes provide information about the array

  • shape

    returns the dimensions of the array as a tuple (

    (3, 4)

    )

  • size

    returns the total number of elements in the array (

    12

    )

  • ndim

    returns the number of dimensions (axes) of the array (

    2

    )
• Access elements using indexing and slicing
  • Use square brackets

    []

    with index or slice notation (

    arr[0]

    ,

    arr[1:5]

    )
  • Use comma-separated indices to access elements in multi-dimensional arrays (

    arr[1, 2]

    )
  • Slice arrays using the

    start:stop:step

    syntax (

    arr[0:6:2]

    )
  • Use fancy indexing to select multiple non-contiguous elements (

    arr[[0, 2, 4]]

    )
• Reshape arrays to change their shape without altering data
  • Use

    reshape()

    to change the shape of an array (

    arr.reshape(2, 6)

    )
  • Flatten multi-dimensional arrays into 1D arrays using

    [flatten()](https://www.fiveableKeyTerm:flatten())

    or

    [ravel()](https://www.fiveableKeyTerm:ravel())

    (

    arr.flatten()

    )
• Join and split arrays
  • Concatenate arrays using

    [np.concatenate()](https://www.fiveableKeyTerm:np.concatenate())

    ,

    [np.vstack()](https://www.fiveableKeyTerm:np.vstack())

    , and

    [np.hstack()](https://www.fiveableKeyTerm:np.hstack())

    (

    np.concatenate((arr1, arr2))

    )
  • Split arrays into smaller arrays using

    [np.split()](https://www.fiveableKeyTerm:np.split())

    ,

    [np.vsplit()](https://www.fiveableKeyTerm:np.vsplit())

    , and

    [np.hsplit()](https://www.fiveableKeyTerm:np.hsplit())

    (

    np.split(arr, 3)

    )

NumPy Functions and Data Processing

NumPy functions for data processing

• Perform mathematical operations on arrays
  • Element-wise arithmetic operations using

    +

    ,

    -

    ,

    *

    ,

    /

    , and

    **

    (

    arr1 + arr2

    )
  • Apply functions like

    [np.sqrt()](https://www.fiveableKeyTerm:np.sqrt())

    ,

    np.exp()

    , and

    [np.log()](https://www.fiveableKeyTerm:np.log())

    for element-wise computations (

    np.sqrt(arr)

    )
  • Use trigonometric functions like

    np.sin()

    ,

    [np.cos()](https://www.fiveableKeyTerm:np.cos())

    , and

    [np.tan()](https://www.fiveableKeyTerm:np.tan())

    on arrays (

    np.sin(arr)

    )
  • Utilize universal functions (ufuncs) for efficient element-wise operations
• Calculate statistical measures
  • Summary statistics using

    [np.mean()](https://www.fiveableKeyTerm:np.mean())

    ,

    [np.median()](https://www.fiveableKeyTerm:np.median())

    ,

    [np.std()](https://www.fiveableKeyTerm:np.std())

    , and

    [np.var()](https://www.fiveableKeyTerm:np.var())

    (

    np.mean(arr)

    )
  • Find the minimum and maximum values using

    [np.min()](https://www.fiveableKeyTerm:np.min())

    and

    [np.max()](https://www.fiveableKeyTerm:np.max())

    (

    np.max(arr)

    )
  • Compute the sum and product of array elements using

    [np.sum()](https://www.fiveableKeyTerm:np.sum())

    and

    [np.prod()](https://www.fiveableKeyTerm:np.prod())

    (

    np.sum(arr)

    )
  • Specify the axis for operations to work along different dimensions (

    np.mean(arr, axis=0)

    )
• Utilize broadcasting to work with arrays of different shapes
  • NumPy automatically broadcasts arrays to make their shapes compatible (

    arr1 + arr2

    )
• Mask and filter arrays
  • Create boolean masks using comparison operators (

    >

    ,

    <

    ,

    ==

    ,

    !=

    )
  • Use boolean masks to filter arrays and select specific elements (

    arr[arr > 5]

    )
  • Combine boolean masks using logical operators (

    &

    ,

    |

    ,

    ~

    )
• Generate random numbers
  • Use the

    np.random

    module to generate random numbers (

    np.random.rand(3, 4)

    )
  • Create arrays with random values using functions like

    [np.random.rand()](https://www.fiveableKeyTerm:np.random.rand())

    and

    [np.random.randn()](https://www.fiveableKeyTerm:np.random.randn())

    (

    np.random.randn(5)

    )
  • Generate random integers using

    [np.random.randint()](https://www.fiveableKeyTerm:np.random.randint())

    (

    np.random.randint(0, 10, size=(2, 3))

    )

Advanced NumPy Concepts

• Structured arrays for heterogeneous data types
  • Create arrays with named fields and different data types
  • Access fields using dot notation or string indexing
• Memory layout and performance optimization
  • Understand row-major (C-style) vs. column-major (Fortran-style) order
  • Use strides to efficiently access array elements in memory
• Array views vs. copies
  • Create views of arrays without copying data
  • Understand when operations create new arrays or modify existing ones