Preliminary Code

Question 1

Why do we need to be concerned with data handling?

Answer 1

Before starting to train a model, we need to get the data into a form which is compatible with the training and testing

Question 2

Why do we train and test in batches?

Answer 2

Training all the full dataset all at once as this would take too long.

Question 3

What is the main aim of building models (in classification or regression)?

Answer 3

To make predictions on unseen data.

This is known as generalisation.

Question 4

What is the square root function?

Answer 4

np.sqrt(num)

Question 5

What is a useful feature of NumPy for scientific computing?

Answer 5

It can deal with sets of numbers in arrays. Arrays only contain with one type, often numbers.

Question 6

How can we create an array?

Answer 6

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

ie passing it a list

Question 7

How do we check the type of an object?

Answer 7

type(object)

Question 8

How do we access elements of a 1D array?

Answer 8

Same indexing as a list
array[1] - single index
array[2:] - every element from position 2 onwards

Question 9

What kinds of arrays are usually used for image data”?

Answer 9

A 2D array is often used for an image, where each element of the array is the value of a pixel in the image.

Question 10

How do you create a 2D array?

Answer 10

Combining multiple 1D arrays with the same size. Enclosing each in a curved bracket and separate with a comma.

array = np.array([(1,2), (3,4)])

Question 11

How can you investigate the size of an array in each dimension?

Answer 11

array.shape

Question 12

How do we tell python what type of 1D vector we want (column or row vector)?

Answer 12

rowVec = array.reshape(1,-1)
rowVec .shape is (1,7)

colVec = array.reshape(-1, 1)
colVec.shape is (7, 1)

Question 13

How can you see how many dimensions are in your array?

Answer 13

array.ndim

Question 14

How can you see the total number of elements in a multi-dimensional array?

Answer 14

array.size

Question 15

How do you index a 2D array?

Answer 15

array[1, 2]

Question 16

What does the : by itself indicate in selection?

Answer 16

You select the whole 1D array along that dimension.

eg array2D([3, :])

Question 17

What are two built-in ways to quickly build arrays?

Answer 17

• linspace()
• arrange()

Both outpace 1D array of numbers

Question 18

What does linspace do?

Answer 18

Outputs evenly spaced numbers between the “start” and “stop” values

np.linspace(start, stop, num=50, endpoint=True, retstep=False, dtype=None, axis=0)

The default number of elements is 50.

Endpoint set to true, last number included by default.

Question 19

What does arange do?

Answer 19

Allows us to define spacing and the length of the array.

arange(start, stop, step , dtype=None)

Default start is 0 and default step is 1.

Question 20

What are linspace and arange functions useful for?

Answer 20

Useful for building iterators.

eg x = np.arrange(0 , 5.1, 0.1)
for n in x:
print(x)

Question 21

What should you check before joining arrays?

Answer 21

Ensure they have the same size along the dimension where you want to join them.

Question 22

What are functions for stacking arrays?

Answer 22

vstack - vertical stacking
hstack - horizontal stacking
dstack - stack in depth

Question 23

What are different ways you can initialise arrays?

Answer 23

• Pass a list
• Create an array of all zeros
• Create an array of all ones
• Create a constant arrays
• Create a random array

Question 24

How do you create an array of zeroes?

Answer 24

np.zeros((2,2))

Question 25

How do you create an array of ones?

Answer 25

np.ones((2,2))

Question 26

How do you create a constant array?

Answer 26

np.full((2,2), 7)

Question 27

How do you create an array of random numbers?

Answer 27

np.random.random((4,4))

Question 28

What is the importance of being able to generate random numbers?

Answer 28

In machine learning random numbers are important. Eg to make an initial guess

Question 29

How do you add/subtract/multiply a matrix by a scalar?

Answer 29

a * 2 a + 2 a - 2

Question 30

How do you perform element-wise addition/subtraction/multiplication/division of matrices?

Answer 30

a+b a-b a*b a/b

Question 31

How do you perform matrix multiplication?

Answer 31

np.matmul(a,b) This is what PyTorch will use. c_ij is the dot product of the ith row of a and the jth column of b.

Question 32

How do we create a replicate of an array, rather than just pointing to the same point in memory?

Answer 32

b = a.copy()

Question 33

How do we import pyplot?

Answer 33

import matplotlib.pyplot as plt

Question 34

How do we plot data?

Answer 34

plt.plot(x, y)

Question 35

How do we add an x or y label?

Answer 35

plt.xlabel("x") plt.ylabel("y")

Question 36

What must we add to the end of each plot?

Answer 36

plt.plot()

Question 37

What do we need to do when plotting two curves on one plot?

Answer 37

Add labels to distinguish the curves. eg plt.plot(x, y, label="sin(x)") And need to add plt.legend()

Question 38

How do we see what curve corresponds to what function?

Answer 38

plt.legend()

Question 39

How do we plot curves of functions?

Answer 39

Generate a set of data points using linspace or arange - set as x. Y is a function of x. Eg: x = np.linspace(0, 20, 1000) y = np.sin(x)

Question 40

How do you customise the line colour and style of a plot?

Answer 40

Add argument "r--" eg plt.plot(x, y, 'r--', label = 'y = sin(x)') "r--" - red dashed line "k-" - solid black line

Question 41

How do you plot two curves side by side?

Answer 41

fig, axs = plt.subplots(1, 2, figsize=(10, 4), tight_layout=True) axs[0].set_xlabel("x", fontsize = 20) axs[0].set_xlim([0, 20]) axs[0].tick_params(axis = 'x', labelsize = 20) axs[0].legend(loc = 'lower right', fontsize = 20)

Question 42

How do we import pandas?

Answer 42

import pandas as pd

Question 43

How do we create a data frame?

Answer 43

From an array: pd.DataFrame(array) From a dictionary: pd.DataFrame(dictionary, index=col_name)

Question 44

How do we make a data series or a data frame from a data frame column?

Answer 44

df[['col_name']] [] - data series [[]] - keep as dataframe

Question 45

How do you select a row by index label?

Answer 45

df.loc[["label"]]

Question 46

How do you load in a dataset from a file?

Answer 46

pd.read_csv('file.csv',index_col=0)

Question 47

How do you plot information in a dataframe?

Answer 47

df.plot(x="col", y="col", kind="scatter")

Question 48

How do you change the scale of a plot?

Answer 48

plt.yscale('log')

Question 49

How do you plot by a categorical variable?

Answer 49

cats = set(df['category'].values) Then can iterate over the categories. Subset the dataframe to have a df of this category. When plotting the categories, use list(cats)

Question 50

How do you extract the values of a column in a dataframe?

Answer 50

.values

Question 51

How do you reduce things to a 1D structure?

Answer 51

.flatten()

Question 52

How do you ensure that two side by side graphs do not overlap?

Answer 52

Add tight_layout=False to plt.subplots()

Question 53

How do you plot with a log scale on the y axis?

Answer 53

ax.semilogy

Question 54

How do you sort values of a data frame?

Answer 54

df.sort_values("col", ascending=False)

Question 55

How can you see the methods and attributes each object has?

Answer 55

dir(object)

Question 56

How do we capitalise a string?

Answer 56

"hello".capitalize()

Question 57

What does string1+string2

Answer 57

concatenates the two strings together

Question 58

What is def __init__(self, vars)?

Answer 58

A special method run each time an instance is created.

Question 59

What is the first argument of any method created within a class?

Answer 59

self This refers to the created instance of the object. This is used to refer to object methods.

Question 60

How do we define a function?

Answer 60

def function(vars):