AWS Machine Learning Foundations Course - Lesson 2

Question 1

What does log loss seek to do

Answer 1

Calculate how uncertain your model is about the predictions it is generating - how likely a model thinks the predictions being generated are to be correct

Question 2

how would you define Hyperparameter

Answer 2

Settings on the model which are not changed during training, but can affect how quickly or how reliably the model trains

Question 3

What does a FFNN do

Answer 3

structures neurons in a series of layers, with each neuron in a layer containing weights to all neurons in the previous layer

Question 4

How would you define discrete

Answer 4

refers to the outcome taking on only a finite number of values like days of the week

Question 5

How would you define data vectorization

Answer 5

Process that converts non-numeric data into a numerical format, so that it can be used by a machine learning model

Question 6

How would you define clustering

Answer 6

Helps to determine if there are any naturally occurring groups in the data

Question 7

How would you define accuracy

Answer 7

The fraction of predictions a model gets right

Question 8

What does CNN represent

Answer 8

Nested filters over grid-organized data

Question 9

What are CNNs most used for

Answer 9

Processing images

Question 10

How would you define a continuous label

Answer 10

Does not have a discrete set of possible values, the label, in theory, could be anything

Question 11

How would you define a categorical label

Answer 11

Has a discrete set of possible values

Question 12

What does bag of words do

Answer 12

Counts how many times a word appears in a document (corpus) and then transforms that information into a dataset

Question 13

How would you define bag of words

Answer 13

technique used to extract features from the text

Question 14

How would you define loss function

Answer 14

measurement of how close the model is to its goal

Question 15

What are some aspects that can ultimately be the largest factor that affects how well you can expect your model to perform

Answer 15

1. Outliers
2. Missing or incomplete values
3. Data that needs to be transformed or preprocessed so its in the correct format to be used by the model

Question 16

What is the fundamental question to ask for data collection

Answer 16

Does the data I have collected match the machine learning task and problem I have defined

Question 17

What are the four aspects of working with data

Answer 17

1. Data collection
2. Data inspection
3. Summary Statistics
4. Data Visualization

Question 18

If your data is categorical, what task would you be working with

Answer 18

Classification

Question 19

If your data is numerical, what task would you be working with

Answer 19

Continuous

Question 20

What is a deep learning model composed of

Answer 20

Collections of neurons connected together by weights

Question 21

What is a tree based model

Answer 21

They learn to categorize or regress by building an extremely large structure of nested if/else blacks, splitting the world into different regions at each block

Question 22

What does training determine in tree based models

Answer 22

Where splits happen and what value is assigned at each leaf region

Question 23

What are linear models good for

Answer 23

Giving a baseline against which to compare more complex models

Question 24

What is a linear model

Answer 24

The relationship between a set of input numbers and a set of output numbers through a linear function

Question 25

What is the end-to-end training process

Answer 25

1. Feed the training data into the model 2. Compute the loss function on the results 3. Update the model parameters in a direction that reduce loss

Question 26

What type of algorithm would you use to segment your customers into multiple groups

Answer 26

If you don’t know how to define the groups, then you can use a cluttering algorithms (unsupervised learning) to segment your clusters of similar customers If you know what groups you would like to have, then you can feed many examples of each group to a classification algorithm (supervised learning) and it will classify all your customers into these groups

Question 27

What type of machine learning algorithm would you use to allow a robot to walk in various unknown terrains?

Answer 27

reinforcement learning

Question 28

Do you want the RMS (root mean square) to be high or low

Answer 28

Low

Question 29

How would you define root mean square (RMS)

Answer 29

Roughly the average error across the test dataset, in general as the model improves, the better the RMS result will be

Question 30

Would you frame the problem of spam detection as a supervised leaning problem or an unsupervised learning problem

Answer 30

Spam detection is a typical supervised learning problem: the algorithm is fed many emails along with their labels (spam not spam)

Question 31

How would you define impute

Answer 31

Refers to different statistical tools which can be used to calculate missing values from the dataset

Question 32

How would you define classification

Answer 32

The process of using machine learning to identify different cases based on patterns found in data (example: spam not spam)

Question 33

What is a common cluster finding model

Answer 33

K - means

Question 34

How old you define training dataset

Answer 34

The data on which the model will be trained

Question 35

What is the result of splitting the dataset

Answer 35

1. A training dataset | 2. A test dataset

Question 36

Why do you split the dataset

Answer 36

To test against the bias variance - trade off

Question 37

What does splitting the data accomplish

Answer 37

Allows you to keep some data hidden during training so that data can be used to evaluate your model before you put it into production

Question 38

Name four unsupervised learning tasks

Answer 38

1. Clustering 2. Visualization 3. Dimensionality 4. Association Rule Learning

Question 39

How much of your data will be in the training dataset after splitting How much of your data will be in the test dataset after splitting

Answer 39

1. Training will be 80% | 2. Test will be 20%

Question 40

How would you define test dataset

Answer 40

The data withheld from the model during training which is used to test how well your model will generalize to new data

Question 41

How would you define unsupervised learning

Answer 41

There are no labels for the training data, the algorithm tries to learn the underlying patterns or distributions that govern the data

Question 42

Name three common unsupervised learning tasks

Answer 42

1. Clustering 2. Association 3. . Dimensionality Reduction

Question 43

What does a transformer do

Answer 43

Enables training over larger datasets involving sequences of data - it is a more modern replacement for RNN/LSTMs

Question 44

What are the two most common supervised tasks

Answer 44

1. Regression | 2. Classification

Question 45

How would you define supervised learning

Answer 45

1. Data is labeled (already has the solution) 2. Every training sample from the dataset 3. Has the corresponding label or output 4. Value associated with it and as a result the algorithm learns to predict labels or output

Question 46

How would you define reinforcement learning

Answer 46

The algorithm figures out which actions to take in a situation to maximize a reward (in the form of a number) on the way to reaching a specific goal

Question 47

What is the difference between supervised and unsupervised learning

Answer 47

Supervised learning uses labeled input and output data, unsupervised learning does not have/use labeled data

Question 48

What are two main identifies for supervised leaning tasks

Answer 48

1. Categorial label | 2. Continuous label

Question 49

How would you define silhouette coefficient

Answer 49

A score from -1 to 1 describing the clusters found during modeling

Question 50

How would you define stop words

Answer 50

A list of words removed by natural language processing tools when building a dataset

Question 51

In silhouette coefficient what does a score loss than zero indicate

Answer 51

Successful identification of discrete non-overlapping clusters

Question 52

Which of the following problem statements fir the definition of a regression - based task 1. I want to detect when my cat jumps on the dinner table, so I set up a camera and write a program to determine if my cat is in the frame or is not in the frame 2. I want to determine the expected reading time for online news articles, so I collect data on my reading time for a week and write a browser plug-in to use that data to predict the reading time for new articles 3. I believe my customers fall into one of many customer segments, but I don’t know what those segments are in advance. After asking for permission, I collect a bunch of data on their actions when they use my reduction and try to determine if there are any collections of users that behave in similar ways 4. I work for a shore company and want to provide a service to help parents predict their children’s show size for any particular age. Within this system, I represent show size as a continuum of values and then round to the nearest show size

Answer 52

2 and 4

Question 53

True or False : Your data requirements will not change based on the machine learning task you are using

Answer 53

False

Question 54

True or False : Models are universal, so the data is not relevant

Answer 54

False

Question 55

True or False: Data needs to be formatted so that it is compatible with the model and model training algorithm you plan to use

Answer 55

True

Question 56

True or False: Data Visualizations are the only way to identify outliers in your data

Answer 56

False

Question 57

True or False : After you start using your model (performing inference), you don’t need to check the new data that it receives

Answer 57

False

Question 58

True or False : The loss function measures how far the model is from its goal

Answer 58

True

Question 59

Why do you need to split the data into training and test data prior to beinginning the model training

Answer 59

If you use all the data you have collected during training, you won’t have any with which to test the model during the model evaluation phase

Question 60

What makes hyperparameters different than model parameters

Answer 60

Hyperparameters are not updated during model training and are set manually

Question 61

True or False: Model evaluation is not very case - specific

Answer 61

False

Question 62

What does model inference involve

Answer 62

1. Generating predictions 2. finding patterns in your data 3. using a trained model 4. testing your model on data it has not seen before

Question 63

True or False : Only a single metric can be used to evaluate a machine learning model

Answer 63

False

Question 64

True or False : The model training algorithm updates a model’s parameters to minimize some loss function

Answer 64

True

Question 65

True or False : Supervised learning uses labeled data while training a model, and unsupervised learning uses unlabeled data while training a model

Answer 65

True

Question 66

In silhouette coefficient what does a score of near zero indicate

Answer 66

overlapping clusters

Question 67

What can RNN/LSTMs be used for

Answer 67

Processing sequences of data

Question 68

What do RNN/LSTMs do

Answer 68

Structured to effectively represent for loops in traditional computing, collecting data while iterating over some object

Question 69

What is a labeled training set

Answer 69

a training set that contains the desired solution(aka label) for each instance

Question 70

How would you define the term labeled

Answer 70

Data that already contains the solution

Question 71

How would you define plane

Answer 71

A mathematical term for a flat surface (like a piece of paper) on which two points can be joined by a straight line

Question 72

How would you define neurons

Answer 72

Simple computational units of neural networks

Question 73

How would you define weights

Answer 73

Mathematical representations of how much information to allow to flow from one neuron to the next The trainable model parameters that are the connections between neural networks

Question 74

Name four neural network structures

Answer 74

1. FNNN 2. CNN 3. RNN/LSTM 4. Transformer

Question 75

How would you define neural networks

Answer 75

Acollection of very simple models connected together

Question 76

What is good tool to use to see outliers and trends in your data

Answer 76

Data Visualization

Question 77

How would you define outliers

Answer 77

Data points that are significantly different from others in the same sample

Question 78

How would you define model parameters

Answer 78

Settings or configurations the training algorithm can update to change how well the model behaves

Question 79

What are some other terms for model parameters

Answer 79

1. Weights | 2. Biases

Question 80

What does a model training algorithm actually do

Answer 80

Iteratively update model parameters to minimize some loss function

Question 81

How are model training algorithms used to train a model

Answer 81

They work through an iterative process where the current model iteration is analyze to determine what changes can be made to get closer to the goal. Those changes are made and the iteration continues until the model is evaluated to meet the goals

Question 82

What is the first step in model training

Answer 82

Randomly split the data set

Question 83

How would you define model inference

Answer 83

When the trained model is used to generate predictions