AI 900 (Masoud) Flashcards
1
Q
What is AI
A
It is basically the concept of human intelligence inside of machines.
2
Q
Artificial General Intelligence
A
When you talk about general A.I., we’re talking about a computer system that can take control of itself. It’s self-aware. It can teach itself a new task, quite a common theme within some science fiction movies. It does not currently exist in real life in any great form.
Azure ====> A.I.
Azure is not general A.I.
3
Q
Narrow Artificial Intelligence
A
Computer systems which use human intelligence but have very strong limitations in what they can do.
such as: Siri, Cortana
4
Q
Machine Learning
A
The study of computer algorithms that improve automatically through experience.
1.Unsupervised learning : ability to find patterns in data without human help
- Supervised learning: humans label the data and give general guidance
5
Q
Natural Language Processing
A
The second area of AI:
Allows a machine to read and understand human language.
Machine translation, question answering, sentiment analysis
6
Q
Perception
A
Third area of AI:
The ability to use input from sensors- images, audio, lidar (light detection and ranging), sonar (Sound Navigation and Ranging), radar (Radio Detection And Ranging), and touch.
Covers things like facial recognition, speech recognition and object recognition.
7
Q
What is ML
A
ML allows computers to use data to forecast the future without specifically being programmed.
8
Q
In ML, a model is a ….
A
is a program that can be used to recognize a pattern in a data
- A model can be used to “predict” future behaviors.
- A model can be used to “categorize” something as one thing or another
- A model can be used to “recognize” people, objects and landmarks using unseen images.
- A model can be used to “understand” the context of natural human text or speech
9
Q
train
A
You “train” a model using “training data”
10
Q
Evaluate
A
You “evaluate” a model using “test data” to measure how accurate is it.
11
Q
Deploy
A
Once a model has been deployed, it can recognize patterns in data it has never seen before
12
Q
Some common AI workloads
A
- Prediction and Demand Forecasting
- Anomaly detection
- Computer Vision
- Natural Language Processing (NLP)
- Conversational AI - Chat Bots
13
Q
Using ML to predict
A
- Give the machine all the relevant data you know
- Tell it for which field you want to predict
- It develops a model which it uses to make a prediction
14
Q
the prevalence of AI
A
causes some ethical and moral challenges
unintended consequences of leaving important decision to a computer
15
Q
Unintended consequences
A
- Decisions that are wrong
- Decisions that are illegal (or at least , go against your own values)
- Decisions that cannot be explained by anybody
- Decisions that are harmful to society at large
16
Q
Six Principles Should Guide AI Development
A
code: “TARIFS “
- Transparency
- Accountability
- Reliability and safety
- Inclusiveness
- Fairness
- Security and Privacy
17
Q
Principle of Fairness
A
AI systems should treat everyone fairly and avoid affecting similarly situated groups of people in different ways.
18
Q
Principle of Reliability and Safety.
A
To build trust, it’s critical that AI systems operate reliably, safely, and consistently under normal circumstances and in unexpected conditions
19
Q
Principle of Privacy and Security
A
Many countries and regions in the world are developing new standards and laws
to try to protect the data of its citizens. Laws are always slower than technology.
20
Q
Principle of Inclusiveness
A
At Microsoft, we firmly believe everyone should benefit from intelligent technology, meaning it must incorporate and address a broad range of human needs and experiences.
21
Q
Principle of Transparency
A
When AI systems are used to help inform decisions that have tremendous impacts on people’s lives, it is critical that people understand how those decisions were made.
22
Q
Principle of Accountability
A
The people who design and deploy AI systems must be accountable for how their systems operate.
23
Q
Common Machine Learning Types
(Regression)
A
A type of supervised learning
The ability to predict the outcome variable
given 1 or more predictor variables.
(Predictor variable, also known sometimes as the independent variable, is used to make a prediction for dependent variables)
The result is numeric - price, amount, size, etc.
It’s about finding the relationships between the variables between the X axis and the Y axis.
24
Q
Common Machine Learning Types
(Classification)
A
A type of supervised learning
Cluster analysis - assign a score to the odds of it belonging to a cluster
What type of fruit is this?
25
Types of Classification
1- Binary classification
only has two answers,0 and 1
2- Multi-class classifications
allow for other options
26
Common Machine Learning Types
(Clustering)
A type of unsupervised learning
Find groups of related things among data
What traits do my best customers
have in common?
27
Core Machine Learning Concepts
1 - Feature:
is an input variable
2- Label:
is the thing we’re predicting
28
Features and Labels in a Dataset
Given a pile of data, you (data scientist) need to determine which bits are relevant
to make decisions on
● Experiment
● Domain knowledge
● Keep in mind the principles of AI
29
Evaluate the Results - Regression
Use the validation dataset to test the model, and measure how close or far the actual results are from the predicted results
"Mean Square Error"
"Large differences" are much "worse" than small differences.
30
Evaluate the Results - Classification
The result is to give a prediction score that the subject is part of the group
“70% confident this is an apple, 30% confident this is a pear”
So if an apple is mis-identified as a pear, that’s ok as long as it only happens 30%
of the time…
31
False Positives vs False Negatives
Compare true positives with false positives and true negatives with false negatives when evaluating the model.
How important is it to you that it never has a false positive?
Accuracy vs precision
32
Azure Machine Learning
“Azure Machine Learning is a cloud-based service
that helps simplify some of the tasks
and reduce the time it takes
to prepare data, train a model,
and deploy a predictive service.”
33
Automated Machine Learning
(AutoML) Steps
1. Identify the problem - classification, regression or time-series forecasting
2. Choose the environment - Python SDK or ML Studio
3. Specify the source and format of the labeled data
4. Configure the compute
5. Configure the AutoML parameters
6. Submit a training run
7. Review the results
34
TPR
True Positive Rate
Also called :
Hit rate
Sensitivity
Recall
TPR = TP/P = TP / (TP + FN) = 1 - FNR
35
FPR
False Positive Rate
Also called:
Fallout rate
Specificity
FPR = FP/ N = FP / (FP + TN) = 1- TNR
36
Curve
TPR and FPR together can be combined onto a curve
37
ROC curve
Receiver Operating Characteristic
منحنی مشخصه عملکرد سیستم یا منحنی عملیاتی گیرنده
38
AUC
Area Under the Curve
گاهی در منحنی ROC به سطح زیر نمودار توجه میشود که به آن AUC یا گفته میشود
the AUC of a classifier is equal to the probability that the classifier will rank a randomly chosen positive example higher than a randomly chosen negative example
39
Azure ML Designer
Drag the dataset onto the designer canvas
Then you can:
- Visualize the data
- Exclude columns that are not going to be useful for training
- Clean rows with missing data from training
- Normalize the data to make it more useful through transformations
40
No code approaches to design model
Automated ML
ML Designer
41
ML Designer
- Drag the dataset onto the designer canvas
- Examine the data visualize it
- In terms of a machine learning algorithm, you can exclude columns that are not going to be useful for training.
- Clean rows with missing data from training
- Normalize the data to make it more useful through transformations
- Drag training models onto the canvas
So unlike the automated ML, you are going to basically be the one to pick the model.
-
42
When you would consider using a computer vision type service?
Computer Vision Features
- Image Classification:
So you are able to feed the image to that service and it will come back with a single term telling you what it identifies from that image.
- Objective Detection
it can identify multiple objects present in an image. It will not only tell you its best guess in terms of what it is but also locate it for you.
- Semantic Segmentation
this is basically Identifying the boundaries of objects.
- Optical Character Recognition:
It's pretty straightforward. It can basically read text on an image. And it's able to extract the actual text from that image.
- Facial Detection and Facial Recognition.
- Form Recognizer
you can then upload usually this is done for images, but it's for forms, for invoices.
And it will basically be able to identify the various fields on the standard forms and extract the data.
43
Computer Vision services in Azure
1. Computer Vision service
2. Cognitive Services
44
Cognitive Services
- Cognitive Services include many other services under one umbrella
- All services require a KEY and an ENDPOINT :
it gives you a single endpoint and it gives you a single key and you can access all of the cognitive services using that single URL if you start using the computer vision service and then you want to start using the natural language processing, those would be separate URLs if you don't use cognitive services.
so remember we said cognitive services is an umbrella across all of this.
Underneath that, you've got "computer vision", "custom vision" and now "face service".
45
Computer Vision Service
- Pre-trained ML model
- Can recognize over 10,000 objects
- Can generate automatic captions for images and tags
- Content moderation for adult, racy or gory content
- Detect faces
- Text recognition
46
Custom Vision Service
- A model that you can build and train
- Classification or object detection
- Upload an existing data set of images and classes
- Publish the model so that you and others can use it
47
Custom Vision
service separates
out....
Custom Vision service separates
out training and prediction into two
resources
48
Face Service
Another type of computer vision is called a face service
Can recognize a human face in an image
Returns the rectangle coordinates of those 1 or more faces
Can recognize celebrities
Needs to be trained on your own data
49
Azure Face Service
- Face Detection
ability to detect gender approximation of age
- Face Verification
To compare one face against a known face, to identify if they belong to the same person.
- Similar Faces
- Face Grouping
Divides a set of unknown faces into groups
based on similarity. All faces of each group are likely to belong to the same person.
Also supports differentiating by another factor such as expression.
- Identify API
you're basically able to put a name to a picture.
50
Form Recognizer Service
- Ideal for invoices and receipts
- Pre-built model or custom model
- Pre-built model recognizes common
receipt formats, English in USA
51
Computer Vision vs Cognitive Services
Computer Vision offers "image recognition and analysis", while Cognitive Services includes a broader range of AI services such as "natural language processing", "sentiment analysis", and "facial recognition".
Cognitive Services is also a little more expensive, but it does offer more features. It’s worth considering both services when choosing the right one for your needs.
52
When should you use Azure Computer Vision?
You should use Azure Computer Vision when you need to extract text or data from an image or when you need to detect and identify objects or facial features in an image. Businesses can also use Azure Computer Vision to determine the sentiment of an image.
53
When should you use Azure Cognitive Service?
You can use Azure Cognitive Services when you need to interpret or understand the meaning of the text or when you need to recognize the natural language.
Organizations can also use Azure Cognitive Services to predict outcomes or recommendations.
When it comes to Cognitive Services, there are a few key areas where they can be applied:
Understanding natural language: Businesses can leverage this feature to understand the text in images and videos better. It can also accurately process customer reviews and feedback.
Face detection and recognition: This is a popular authentication feature used to track customers and staff. This AI-powered tool can also accurately measure customer sentiment.
Object recognition: Object recognition is probably the most straightforward task for smart algorithms. This AI-powered feature is popular in retail as it can accurately identify products in images and videos.
Speech recognition: If you use Siri or Alexa, you know that speech recognition has come a long way in recent years. Today, companies can use it to transcribe speech into text, optimizing customer experiences.
The great thing about Azure Cognitive Services is that they’re all fully cloud-based and require no additional infrastructure or setup. You can get started with them in minutes, making them an excellent choice for businesses of all sizes.
54
NLP Workload Scenarios
Understands
written and spoken
language
55
Key Phrase Extraction
Identifying the main points of a document; context
56
NLP services types
- Key Phrase Extraction
- Entity Recognition
- Sentiment Analysis
- Language Modeling- Speech Recognition and Synthesis
- Translation
57
Entity Recognition
Identifying “entities” of a document; items categorized by type and subtype.
Another interesting feature of the same entity recognition as it can basically find the entity on Wikipedia.
57
Sentiment Analysis
Pre-built ML model
Sentiment score from 0 to 1 - 1 being positive sentiment
So if I say I love cheese, then that is an extremely positive sentiment.
If I say cheese is okay, it is a somewhat positive sentiment.
If I don't like cheese, which I don't know why you wouldn't, that would be a negative sentiment and something in between.
58
Language Modeling
Build your own dictionary for terms in your industry
Train the language understanding engine for your own use
59
Speech Recognition and Synthesis
Speech recognition - the ability to detect and interpret speech
Speech synthesis - the ability to generate spoken output
Speech-to-text and text-to-speech
60
Translation
Supports over 60 languages
Getting better all the time
Understands semantic context
61
NLP Workload Tools
Microsoft Azure cognitive services for natural language processing are broken up into several services:
- Language Service (fmr Text Analytics)
- Language Service, formerly
Language Understanding Service (LUIS)
- Speech Service
- Translator Text Service
62
Language Service (fmr Text Analytics)
All of the Azure services that include key phrase extraction, entity detection, and
sentiment analysis
It can detect the language of text
Can detect when text has multiple languages and identify the predominant
language
Returns NaN when it can’t determine
63
Language Service, formerly
Language Understanding Service (LUIS)
Understanding natural language
Three core concepts:
1. Utterances
2. Entities
3. Intents
64
Language Service - Utterances
An example of something a user will say:
“What time are you open until?”
“When do you close?”
“What time do you close today?”
“What is the latest you are open until?”
65
Language Service - Entities
An item to which an utterance refers:
“What time do you close today?”
● Today (DateTime)
● You (subject)
66
Language Service - Intents
The purpose or goal expressed by the user
Things your application is able to do
“What time do you close today?”
Goal: TodayHoursOfOperation
67
Language Service - Intents
Sometimes, there is no intent
Questions you don’t even want your AI to try to answer.
“What is the meaning of life?”
“How do teach a dog to walk himself?”
68
Speech Service
Besides that, the language understanding, you also have speech services like :
Text to speech
Speech to text
Audio translation
“Custom voice models”
69
Translator Text Service
Translate text between 70 languages
Custom models for specific industries / terms
Can detect the language being passed in
70
Conversational AI Solutions
Now, conversational AI is basically not only understanding what you're saying, the natural language processing element, but being able to respond to that in a natural way.
Conversational AI allows a human and
a computer to talk
Typically called agents, or bots
71
Azure Bots can operate....
... over the web, email, social
media, and voice
72
Conversational AI Use Cases
- Webchat Bots
- Telephone Voice Menus
- Personal Digital Assistants
73
Webchat Bots
● Appear on a website or social media app
● Understands and responds to written text
● A limited scope of conversation
● Often friendly tone, called “chit chat”
74
Telephone Voice Menus
Speech recognition and speech synthesis
Conversational banking
75
Personal Digital Assistants
OK Google - when is my next appointment?
Hey Alexa - what time is the movie tonight?
Yo yo, Bing - tell Mom that I’ll be home this weekend from college
76
Conversational AI Services
- QnA Maker Service
- Azure Bot Service
77
QnA Maker Service
Automatically build a chat bot based on any semi-structured source
- Your website
- FAQ
- Product manuals
etc
78
Azure Bot Service
Enterprise-grade chat bots
Start simply and grow in sophistication
Coded chat bots
79
What Azure AI Services Exist?
Vision Services
Speech Services
Language Services
Decision Services
80
AKS and ACI
In Azure Machine Learning, you have the option of deploying a predictive service to Azure Container Instance (ACI) or Azure Kubernetes Service (AKS). For production scenarios, you should use an AKS deployment, which requires creating an inference cluster compute target. ACI-based deployment is suitable for testing. Azure Machine Learning does not support deployment of predictive services to Azure Functions or Azure Logic Apps.
