Numpy - Python for Data Science

Python for Data Science & Machine Learning Bootcamp
notes by Lucas, Sep 15, 2018

Numpy

1. Introduction

• The core of Numpy library is Linear Algebra. So Numpy is very fast, and many scientific libraries are based on it;
• Importing Numpy: import numpy as np;

2. Numpy Arrays

• The dimension of nd.array can be seen from the numbers of [...] in array(); e.g. array([...]) indicates 1 dim, array([[...]...]) indicates 2 dim, etc.;
• Note the difference between np.arange() (NOT np.range) and np.linspace();
• np.eye() creates an Identity Matrix;
• np.random.rand() generates the random numbers from 0 to 1;
• np.random.randn() generates the random numbers which obey the normal distribution (mean = 0);
• np.random.randint() generates ONE random number in [start, end), add a third paramter if wanna generates a certain mount of random numbers;
• arr.reshape() creates a specific shape of n dimentional array;
• Note the difference between arr.max() and arr.argmax(), for the real value and for the position (index);
• arr.shape checks the dim of an array;
• arr.dtype checks the data type in an array;
• For np.random.randint(), you can just input from numpy.random import randint, and then just input randint();

3. Numpy Array Indexing & Selection

• Slice: arr[0:5] selects the group of elements whose index are from 0 to 4;
• Broadcast: arr[0:3] = 520 makes the first three elements become a bunch of numbers of 520;
• Copy: arr.copy() makes a copy of the array;
• arr_2d[0][0] equals arr_2d[0,0], but the later notation is recommended;
• Conditional selection: arr[arr>5];
• Combination of functions: arr_2d = np.arrange(0:50).reshape(5,10);

4. Numpy Operations

• arr + arr means every elements of arr has been doubled, and so were the subtraction and multiplication;
• The operation of a numpy array with a scalar number is easy to understand;
• Universal functions: np.sqrt(), np.exp(), np.max(arr) equals arr.max(), np.sin(), np.log();