CPSC150A
Scientific Computing

Creating Numpy Arrays

• Numpy is a module that makes working with multidimensional list easier
• Numpy defines it’s own type called array

• A numpy array is a multidimensional lists where:

  1. all values have the same type
  2. all sub-lists are the same length

• Numpy arrays can be created from lists

  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])

• A couple of differences to note about numpy arrays

• They print differently than lists

  a_list = [[1, 2, 3], [4, 5, 6]]
  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])
  print(a_list)
  print(an_array)

• They also have a different type

  a_list = [[1, 2, 3], [4, 5, 6]]
  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])
  print(type(a_list))
  print(type(an_array))

• Can also create numby arrays filled with a single value

  all_zeros = numpy.zeros((2, 3))
  print(all_zeros)
  all_ones = numpy.ones((2, 3))
  print(all_ones)

• Note the input to the functions specifie the number of rows and columns in the array
• Also note the extra set of parentheses, the function has a single input, an immutable list of the dimensions of the array

• Can also create vectors (one dimensional arrays)

  a_vector = numpy.arange(6)
  print(a_range)

• arange works just like range, you can give it a start, stop, and step

• Unlike range, arange can use floating point numbers

  a_vector = numpy.arange(1, 10, 0.5)
  print(a_vector)

• However, because of floating point precision it is difficult to predict how many numbers will be in the range

• As an alternative you can use linspace

  a_vector = numpy.linspace(1, 10 19)

• The inputs to linspace are start, stop, and num

• Can also turn a vector into a matrix by using a method

  a_vector = numpy.arange(7)
  a_matrix = a_vector.reshape(2, 3)
  print(a_matrix)

• Can also create an array of random values

  an_array = numpy.random.random((2, 3))
  print(an_array)

Numpy Array Attributes

• Once you have a numpy array, you can get information about the array using attributes
• Attributes are information attached to a python object

• Access the information by using the dot notation, like a function, but without the parentheses of a function

  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])
  print(an_array.ndim)
  print(an_array.shape)
  print(an_array.size)
  print(an_array.dtype)

Operations

• Can perform arithmetic operations on two arrays of the same size

  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])
  all_ones = numpy.ones((2, 3))
  sum = an_array + all_ones
  print(sum)

• This performs an elelment-wise operation
• That is each element is matched with the correspoding element in the other array to perform the operation
• Can perform any of the arithmetic or logical operators, +, -, *, /, //, **, <, >, etc.

• This can be used to create an array of any single value

  all_tens = numpy.ones((2, 3)) * 10
  print(all_tens)

• Can also use some functions that are like the math module functions

  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])
  print(numpy.sqrt(an_array))
  print(numpy.sin(an_array))

• Don’t need to import another module to use them, they are in the numpy module

• Can also perform arithmetic operations on an array and a single value

  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])
  sum = an_array + 1
  print(sum)

• This has the effect of applying the same operation to every element in the array

• Can also use methods

  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])
  print(an_array.sum())
  print(an_array.min())
  print(an_array.max())

• These methods ignore the structure of the array and find the sum, min, and max of all the elements

• It is possible to perform these along particular axes using an optional parameter

  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])
  print(an_array.sum(axis=0))

• This sums along the first axis, or the rows in this example

Indexing and Slicing

• One dimensional arrays can be indexed and sliced just like lists

  a_list = [1, 2, 3, 4, 5, 6]
  an_array = numpy.array([1, 2, 3, 4, 5, 6])
  print(a_list[2])
  print(an_array[2])
  print(a_list[1:3])
  print(an_array[1:3])

• Multi dimensional arrays are indexed and sliced differently than lists

  a_list = [[1, 2, 3], [4, 5, 6]]
  an_array = numpy.array([[1, 2, 3], [4, 5, 6]])
  print(a_list[1][2])
  print(an_array[1, 2])
  print(an_array[1:3, 0])

Fancy Indexing

• Can also index using an array of indices

  an_array = numpy.array([1, 2, 3, 4])
  an_array[[0, 2]] = 0
  print(an_array)

• Or with an array of booleans

  an_array = numpy.array([1, 2, 3, 4])
  an_array[[True, False, True, False]] = 0
  print(an_array)

Activity