Azure AI-900 - Notes!

Hello Friends,

Last week, I mentioned that we are facing some challenges while deploying the synapse artifacts to higher environment, still the issue is not resolved. I will definitely post the issue and solution, as we have spent almost 2-3 weeks on this.

This week, I thought to share my preparation notes on Exam AI-900. Hope this helps to those who are planning for writing this exam in near future.

Lets start with some basics:

What is AI?

As per oxford dictionary, the noun word “artificial intelligence” means the study and development of computer systems that can copy intelligent human behavior.

So, basically it is incorporating human intelligence inside the machines, having strong limitations in what they do. We can even categorize it under “Narrow artificial intelligence”. Go and checkout "what is general aritifical intelligence"!

What is machine learning?

Machine learning is the subset of AI, and it is mainly the study of computer algorithms that improve automatically via experience.

There are two types of ML:

1. Unsupervised learning – the ability to find patterns in data without human.
2. Supervised learning – humans label the data and gives the general guidance.

What is NLP?

Its another field of AI, allows machines to read and understand the human language.

What is perception?

This is another field of AI, that uses inputs from sensors such as images, audio, sonar, radar etc. Covers things like facial recognition, speech recognition & object recognition.

As per Microsoft documentation, I am going to divide this into four major sections-

Section#1:

Describe Artificial Intelligence workloads and considerations (20–25%)

This section covers the features of some of the common AI workloads and its guiding principles.

Common AI Workloads:

Below are the list of common AI workload types->

Workload Type#1: Prediction & Demand Forecasting

This type of workload uses machine learning to predict.

1.??????Provide data to predict.

2.??????Identify field you want the prediction.

3.??????ML develops model which it uses for prediction.

Common example are Netflix recommendations; amazon/other e-shopping sites etc...

Workload Type#2: Anomaly Detection

This type of workload uses machine learning to predict-

1.??????Data points or observation that is different from rest of the data points or observation.

2.??????It's also called outlier detection.

Real-time example can be – credit card company can use anomaly detection to track how customers use their credit cards. Other applications like health monitoring or other finaniancial & banking institutions.

Workload Type#3- Computer Vision

This type of workloads predicts meaningful information from the visual inputs provided for the prediction.

Common example are facial detection, self-driving cars, gaming & fitness apps.

Workload Type#4- Natural processing language

Predict what human language is asking for.

Part of NLP is Knowledge Mining (Cognitive search)- which get content from blob storage, SQL table, cosmos DB or SQL DB or pdf, word documents... etc. Based on these it creates a searchable index.

Example- google search engine.

Workload Type#5-Conversational AI

Chat bots comes under this type of workload type. Related to NLP but turning it into application and hook it up with any chat technology that it supports. It's powered by AI.

Guideline of AI:

The six principles that should guide AI development are summarized below-

1.??????Fairness

“AI systems should treat everyone fairly and avoid affecting similarly situated groups of people in different ways.”

2.??????Reliabilty & Safety

“To build trust, it's critical that AI systems operate reliably, safely, and consistently under normal circumstances and in unexpected conditions.”

3.??????Privacy & security

“Bring up new laws to protect the data of its citizens.”

4.??????Inclusiveness

“It must incorporate and address a broad range of human needs and experiences.”

5.??????Transparency

“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.”

6.??????Accountability

“The people who design and deploy AI systems must be accountable for how their systems operate.”

Section#2:

Describe fundamental principles of machine learning on Azure (25- 30%)

You need to be familiar with the common machine learning types. For this exam, high-level would be sufficient.

Common machine learning types:

1.??????Regression- supervised learning. Finds the relationship between the variables.

2.??????Classification- supervised learning. Binary classification (0 or1) or multi-class classification.

3.??????Clustering- unsupervised learning. Find groups of related things among data.

Core- machine learning concepts:

1.??????Feature – an input variable

2.??????Label – thing we are predicting

Example of feature & label in the given set of data-

First and last name ● Smoker or non-smoker (feature) ● Date of birth (feature) ● Marital status (feature) ● Gender (feature) ● Annual income ?● Blood type ● Number of children ● … ● Life expectancy (label)

Once you have the feature and the label identified. We need to follow below steps-

Train the model

Some important points while training the model->

1.??????The more input (historical) data you have, the more accurate the results.

2.??????Don’t use ALL of your data to train the model.

3.??????Divide your available labelled data into training and validation/test datasets.

Split the training & validation data randomly

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.

Note: Mean Square Error - Large differences are much worse than small differences.

Evaluate the results – classification-

The result is to give a prediction score that the subject is part of the group.

Compare true positives with false positives and true negatives with false negatives when evaluating the model

Understand the true positive, false positive, trues negative & false negative concepts.

Capabilities of no-code ML

I. AutoML

Understand the concept of AutoML(Automates Machine learning). Below are the high-level 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.

II. Azure ML designer

Azure ML designer is another no code ML method. Below are the high-level steps for designing in ML designer-

1. Drag the dataset onto the designer canvas.
2. Visualize the data.
3. Exclude columns that are not going to be useful for training.
4. Clean rows with missing data from training.
5. Normalize the data to make it more useful through transformations.
6. Drag training models onto the canvas.
7. Training pipeline can be converted to the inference pipeline.

Section#3:

Describe features of computer vision workloads on Azure (15-20%)

Computer Vision Features

1.??????Image classification

2.??????Object detection

3.??????Optical character recognition

4.??????Facial detection & recognition

Computer Vision Tools

In Azure, we can use either computer vision service or cognitive service. Cognitive services includes many other services under one umbrella.

Please note- all services requires a key and endpoint.

• 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.

• 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.

Custom Vision service separates out training and prediction into two resources

1.??????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.

2.??????Form Recognizer Service

Ideal for invoices and receipts. Pre-built model or custom model. Pre-built model recognizes common receipt formats.

Section#4:

Describe features of NLP workloads on Azure (25-30%)

This section covers NLP workload scenarios and tools.

NLP Workload Scenarios

Understands written and spoken language.

1.????Key Phrase Extraction: Identifying the main points of a document; context.

2.????Entity Recognition: Identifying “entities” of a document; items categorized by type and subtype.

3.????Sentiment Analysis: Pre-built ML model Sentiment score from 0 to 1 - 1 being positive sentiment.

4.????Language Modeling: Build your own dictionary for terms in your industry Train the language understanding engine for your own use.

5.????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.

6.????Translation: Supports over 60 languages Getting better all the time Understands semantic context.

NLP Workload Tools

1.????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 tex. Can detect when text has multiple languages and identify the predominant language. Returns NaN when it can’t determine.

2.????Language Service, formerly Language Understanding Service (LUIS) Understanding natural language. Three core concepts: 1. Utterances 2. Entities 3. Intents

• Language Service - Utterances: An example of something a user will say. For example: what time will the store be open today?

Today (Date time) in above example.

• Language Service - Intents: The purpose or goal expressed by the user. For example: Goal: TodayHoursOfOperation

Sometimes, there is no intent Questions you don’t even want your AI to try to answer. Example – what is the meaning of life?

3.????Speech Service: Text to speech, Speech to text , Audio translation “Custom voice models”.

4.????Translator Text Service: Translate text between 70 languages; Custom models for specific industries / terms; Can detect the language being passed in.

Conversational AI Solutions & services

This is the last topic, very common nowadays, where Conversational AI allows a human and a computer to talk. Typically called agents or bots. It can operate over the web, email, social media, and voice. Example – Webchat bots, telephone voice menus, Personal digital assistants.

Conversational AI services

1.????Azure Bot Services: Enterprise-grade chat bots; Start simply and grow in sophistication; Coded chat bot.

Conclusion:

I have tried to provide a nutshell of the entire exam content of AI-900. Apart from this, I would suggest checking on Microsoft exam guide and learning path. Also, try to do some practice test on this.

Exam AI-900: Microsoft Azure AI Fundamentals - Certifications | Microsoft Learn

I will let you know once I have passed this certification.

Enjoy learning and sharing ??

Thank You All ??