AWS Machine Learning Foundations Course - Lesson 3 Flashcards

1
Q

What is a machine learning task

A

The outputs of model training algorithms are classified based on the task they are designed to solve

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
2
Q

How do you determine which machine learning task to use

A

The absence or presence of labeling in the the data is used to identify the machine learning task to use

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
3
Q

What are the three components of machine learning

A
  1. A machine learning model
  2. A model training algorithm
  3. A model inference algorithm
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
4
Q

What are the five machine learning steps

A
  1. Define the problem
  2. Build the dataset
  3. Train the model
  4. Evaluate the model
  5. Deploy/use the model
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
5
Q

What are the four main machine learning tasks

A
  1. Supervised
  2. Unsupervised
  3. Semi - Supervised
  4. Reinforcement learning
How well did you know this?
1
Not at all
2
3
4
5
Perfectly
6
Q

How do you define machine learning

A

modern software development technique that enables computers to solve problems by using examples of real world data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
7
Q

What is the difference between a model parameter and a hyperparameter

A

Parameters are set and changed while hyperparameters are parameters of the algorithm itself, not of the model and cannot be changed

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
8
Q

How would you define action

A

For every state, an agent needs to take an action toward achieving its goal

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
9
Q

How would you define agent

A

The piece of software you are training is called an agent. It makes decisions in an environment to reach a goal

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
10
Q

How would you define discriminator

A

A neural network trained to differentiate between real and synthetic (fake) data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
11
Q

How would yo define discriminator loss

A

Evaluates how well the discriminator differentiates between real and fake data

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
12
Q

How would you define edit event

A

When a note is either added or removed from your input track during inference

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
13
Q

How would you define environment

A

The environment is the surrounding area within which the agent interacts

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
14
Q

Describe exploration vs. exploitation

A

An agent should exploit known information from previous experiences to achieve higher cumulative rewards, but it also needs to explore to gain new experiences that can be used in choosing the best actions in the future.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
15
Q

How would you define generator

A

A neural network that learns to create new data resembling the source data on which it was trained

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
16
Q

How would you define generator loss

A

Measures how far the output data deviates from the real data present in the training dataset

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
17
Q

How would you define hidden layer

A

A layer that occurs between the output and input layers. Hidden layers are tailored to a specific task.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
18
Q

How would you define input layer

A

The first layer in a neural network. This layer receives all data that passes through the neural network

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
19
Q

How would you define output layer

A

The last layer in a neural network. This layer is where the predictions are generated based on the information captured in the hidden layers

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
20
Q

How would you define piano roll

A

A two dimensional piano roll matrix that represents input tracks.

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
21
Q

In a piano roll, how are time and pitch resented

A

Time is on the horizontal axis, and pitch is on the vertical axis

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
22
Q

How would you define reward

A

Feedback is given to an agent for each actions it takes in a given state (as a numerical reward)

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
23
Q

What is the AWS machine learning mission

A

To put machine learning in the hands of every developer

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
24
Q

What is Amazon SageMaker

A

A fully managed service that removes complexity from machine learning workflows so every developer and data scientist can deploy machine learning for a wide range of use cases

How well did you know this?
1
Not at all
2
3
4
5
Perfectly
25
What can you accomplish by using AWS pre-trained AI services
You can apply ready-made intelligence to a wide range of applications such as personalized recommendations, modernizing contact center, improving safety and security, and increasing customer engagement
26
How would you define AWS DeepLens
A deep learning enabled video camera that allows you to deploy trained models directly to the device
27
What are four key components required for an WS DeepLens based project
1. Collect your data and store it in an Amazon S3 bucket 2. Train your model using jupiter notebook in SageMaker 3. Deploy your model using AWS Lambda to deploy the trained model to your AWS DeepLens device 4. View model output using Amazon IoT Greenrass after it is deployed
28
How would you define DeepRacer
An autonomous race ca designed to test reinforcement learning models by racing on a physical track
29
How would you define AWS DeepComposer
A composing device powered by generative AI that creates a melody that transforms into a complexity original song
30
What are the three main components of neural networks
1. Input layer 2. Hidden layer 3. Output layer
31
How would you define image classification
The most common application of computer vision in use today. It can be used to answer questions like “whats in this image?” This type of task has application in text detection or optical character recognition (OR) and content
32
How would you define object detection
Closely related to image classification, but allows users to gather more granular detail about an image.
33
How would you define semantic segmentation
Another common application of computer vision that takes a pixel approach. Instead of just identifying whether an object is present or not, it tries to identify, down to the pixel level, which part of the image is part of the object
34
How would you identify activity recognition
An application of computer vision that is based around videos rather than just image. Video has the added dimension of time and, therefore, models are able to detect changes that occur over time
35
In reinforcement learning, what is an agent trained to do
Achieve a goal based on the feedback it receives as it interacts with an environment
36
In reinforcement learning, what does the agent collect
A number as a reward for each action it takes
37
In reinforcement learning, what are the actions that help the agent achieve its goal incentivized with
Higher numbers
38
In reinforcement learning, what do unhelpful actions result in
A low reward or no reward
39
In reinforcement learning what does the agent learn over time (through trial and error)
To map gainful actions to situations
40
What is reinforcement learning particularly useful for
Addressing sequential problems with long term goals
41
Name four real-world examples of reinforcement learning
1. Industrial robotics 2. Fraud detection 3. Stock trading 4. autonomous driving
42
How would you define state
Defined by the current position within the environment that is visible, or known, to the agent
43
How would you define episode
Represents a period of trial and error when an agent makes decisions and gets feedback from its environment
44
How would you define algorithm
A set of instructions that tells a computer what to do
45
How would you define a soft actor critic (SAC)
Embraces exploration and is data - efficient, but can lack stability
46
How would you define a proximal policy optimization (PPO)
Stale but data hungry
47
How would you define an action space
The set of all valid actions, or choices, available to an agent as it interacts with an environment
48
How would you define a discrete action space
Represents all of an agents possible actions for each state in a finite set of steering angle and throttle combinations
49
How would you define a continuous action space
Allows the agent to select an action from a range of values that you define for each state
50
How would you define learning rate
A hyperparameter that controls how many new experiences are counted in learning at each step
51
What does a higher learning rate result in
Faster training, but may reduce the model’s quality
52
What is the purpose of the reward function
To encourage the agent to reach its goal
53
How does the reward function achieve its goal
Each stat on the grid is assigned a score, you incentivize behavior that supports your goal
54
In reinforcement learning, what is one of your most important jobs
Figuring out how to reward which actions
55
How would you define reward function
The actual code you write to elf your agent determine if the action it just took was good or bad, and how good or bad it was
56
How would you define exploration
The more training an agent gets, the more it leans about an environment. This experience helps it become more confident about the actions it chooses. An agent needs to explore to gain new experiences that can be used in choosing the best actions in the future
57
How would you define exploitation
Using information from previous experiences to help it reach its goal. An agent should exploit known information from previous experiences to acheieve higher cumulative rewards
58
How would you define a reward graph
Where the DeepRacer metrics are displayed
59
What does plotting the total reward from each episode allow you to see in a reward graph
how the model performs over time. The higher the reward, the better the model performs
60
How does a balance of exploration and exploitation help a reinforcement learning model
The more an agent learns about its environment the more confident it becomes about the actions it chooses. If an agent doesn’t explore enough, it often sticks to information its already learned even if this knowledge doesn’t help the agent achieves it goal. The agent can use information from previous experiences to help it make future decisions that enable it to reach its goal .
61
In a reward graph, what is represented in the average reward
The average reward the agent earns during training iteration
62
IN a reward graph - how is the average in the average reward calculated
by averaging the reward earned across all episodes in the training iteration
63
In a reward graph, what is represented in the average percentage completion (training)
The average percentage of the track completed by the agent in all episodes run during the evaluation period
64
In a reward graph, what does the best model line allow you to see
Which of your model iterations had the highest average progress during the evaluation
65
In a reward graph, in the best model line, what is a check point
A snapshot of a model that is captured after each training (policy-updating) iteration
66
In a reward graph, what does the reward primary y - axis show
The reward earned during a training iteration
67
In a reward graph, what does the percentage track complete ion secondary y - axis show
The percentage of the track the agent completed during a training iteration
68
In a reward graph, what does the iteration x-axis show
The number of iterations completed during your training job
69
In AWS DeepRacer, when does overfitting or overtraining become an issue
When a model is trained on a specific track for too long
70
What should a good model be able to make decision based on
Features of the road, such as sidelines and center lines, and be able to drive on just about any track
71
What does an over trained model learn to navigate
Landmarks specific to an individual track
72
How would you define discriminative model
It aims to answer the question “if i’m looking at some data, how can i best classify this data or predict a value”
73
How would you define generative model
It aims to answer the question “have i seen data like this before”
74
What can the patterns learning in a generative model be used for
To create brand new examples of data which look similar to the data it saw before
75
How would you define auto regressive convolutional neural networks (AR-CNNs)
Used to study systems that evolve over time an assume that the likelihood of some data depends only on what has happened in the pas.
76
What are AR-CNNs useful for
As a way of looking at many systems, from weather prediction to stock prediction
77
How would you define generative adversarial networks (GANs)
A machine learning model format that involves pitting two networks against each other to generate new content
78
In a generative adversarial network, what does it swap back and forth between and why
A generator network (responsible for producing new data) and a discriminator network ( responsible for measuring how closely the generator networks data represents the training dataset)
79
What are transformer based models most often used for
To study data with some sequential structure (such as the sequence of words in a sentence)
80
What can you use for an input track in DeepComposer
1. A sample track 2. A recorded custom track 3. Imported track
81
In DeepComposer, what can you use for the machine learning technique
Either a sample model or a custom model
82
What are the three supported generative AI techniques in DeepComposer
1. Generative adversarial networks(GANs) 2. Auto-regressive convolutional neural networks (AR-CNNs) 3. Transformers
83
In DeepComposer, what is the GAN technique used for
To create accompaniment tracks
84
In DeepComposer, what is the AR-CNN technique used for
To modify the notes in your input track
85
In DeepComposer, what is the transformers technique used for
To extend your input track by up to 30 seconds
86
In a GAN, what are the two neural networks that are pit again each other
1. a generator | 2. a discriminator
87
How would you define a generator
A neural network that learns to create new data resembling the source data on which it was trained
88
How would you define discriminator
A neural network trained to differentiate between real and synthetic (fake) data
89
What does the generator learn to do
Produce more and more realistic data
90
What does the discriminator do
Iteratively gets better at learning to differentiate real data from the newly created data
91
How would you define generator loss
Measures how far the output data deviates from the real data presenc in the training dataset, this feedback is then used by the generator to incrementally get better at create it realistic music
92
How would you define discriminator loss
Evaluates how well the discriminator is differentiating between real and fake data
93
How is time represented on each two dimensional piano roll
On the horizontal axis
94
How is pitch represented on each two dimensional piano roll
On the vertical axis
95
How would you define edit event
When a note is either added or removed from your input track during inference
96
How would you define binary classifier
It classifies inputs into two groups - “real” and “fake”
97
To detect both the cat and the dog present in an image, what kind of computer vision model would you use
Object detection
98
Which computer vision based task would you use to detect that the dog in an image is sleeping
Image classification
99
Which Computer vision based task would you use to detect the exact location of the car and the dog in an image
Semantic (instance) segmentation
100
What kind of computer vision model would you use to count the number of cars in an image
Object detection