Getting Started with Azure Artificial Intelligence

Introduction

Artificial Intelligence (AI) is revolutionizing the way we interact with technology, enabling machines to mimic human behaviour and cognitive abilities. Azure, Microsoft’s cloud computing platform, offers a powerful suite of AI tools and services that empower developers and data scientists to create cutting-edge AI solutions. In this blog, we will delve into the world of Azure AI and explore some of its key components and functionalities.

What is AI?

At its core, AI involves the development of software that can imitate human behaviour and cognitive processes. Azure AI encompasses various subdomains, including:

Machine Learning

What is Machine Learning?

Machine Learning serves as the backbone of AI. It involves training computers to make predictions and decisions based on vast amounts of data. Instead of explicitly programming a computer to perform a task, it learns from examples and improves its performance over time.

Example: In the realm of healthcare, Machine Learning algorithms can analyze patient data to predict the likelihood of certain diseases, enabling early intervention and personalized treatment plans.

How does Machine Learning work?

Machine Learning works on large amounts of data which is generally generated by common people through text messages, photos, social media posts etc. Data sources can also include data collected from different sensors used in vehicles, factories, cities and even homes.

The computer uses all this data and tries to make predictions and inferences using a Machine Learning model created by data scientists. Following are the steps in a Machine Learning process:

1. Data Collection
2. Data Labelling
3. ML Model trained on labelled data
4. Testing of results
5. Using ML model for prediction or making a conclusion

Microsoft Azure Machine Learning Services

1. Automated Machine Learning: This service democratizes Machine Learning by enabling non-technical users to build efficient ML models effortlessly. eg. A marketing professional can use Automated ML to predict customer churn based on historical sales data without extensive coding knowledge.
2. Azure Machine Learning Designer: Designed as a graphical, low-code platform, the Azure ML Designer allows data scientists to visually construct Machine Learning solutions. eg. Data scientists can use the Azure ML Designer to create a pipeline that automates image classification tasks using convolutional neural networks.
3. Data and Compute Management: Azure provides cloud-based data storage and computing resources, allowing professionals to conduct large-scale data experiments with ease. eg. A data engineer can leverage Azure’s resources to process and analyze big data generated from IoT devices.
4. Pipelines: Pipelines facilitate the seamless orchestration of model training, deployment, and management tasks. eg. An IT professional can create an end-to-end pipeline that automates the deployment of a sentiment analysis model into a web application.

Anamoly Detection

What is Anamoly Detection?

Anamoly detection is a Machine Learning technique which helps in identifying unusual behaviour by analysing data over a time peiod.

Example: Credit Card Fraud Detection

Scenario: Let’s consider a financial institution that issues credit cards to its customers. Each day, thousands of credit card transactions take place, with the majority being legitimate transactions made by cardholders for their purchases. However, there might be some fraudulent transactions carried out by unauthorized users attempting to use stolen credit card information. These can be detected using Anamoly Detection.

Steps for Anamoly Detection:

1. Data Collection: The financial institution collects transaction data from its credit card users. This data includes information such as transaction amount, location, time, merchant, and more.
2. Data Preprocessing: Before applying anomaly detection techniques, the data is preprocessed to remove noise, handle missing values, and normalize the features for better analysis.
3. Training Period: During the training period, the algorithm learns from historical transaction data that includes both legitimate and fraudulent transactions. It analyzes patterns and establishes a baseline for what is considered “normal” behavior.
4. Anomaly Detection: Once the model is trained, it can predict whether a new transaction is legitimate or fraudulent based on the patterns it learned during training. If the new transaction deviates significantly from the established normal behavior, it will be flagged as an anomaly.
5. Real-time Monitoring: As new transactions occur, the model continuously evaluates them in real-time, identifying potential anomalies in the credit card transactions.
6. Alerts and Actions: If an anomaly is detected, the financial institution can take immediate action to prevent fraud. This might involve notifying the cardholder, blocking the card, and initiating a verification process to ensure the legitimacy of the transaction.

Computer Vision

It is the capability of AI which helps it to interpret the world using images, videos and computer cameras.

Computer vision capabilities

1. Image Classification: It involves training AI model to classify an images based on contents.
2. Object Detection: It involves training AI model to identify objects in a image/video and provide a boundary for it to track it’s location.
3. Semantic Segmentation: It is an advanced AI technique which involves Machine Leraning model which analyze each pixel of the image or video and classify based on item they belong to.
4. Image Analysis: This capability allows the computer to add tags or description to an image based on the what it sees in the image.
5. Face Detection: It can used to detect all the faces present in a imge or video using the facial geometry.
6. Optical Character Recognition: This is used to detect or read text in an image.

Microsoft Azure Computer Vision Services

1. Computer Vision: This Microsoft Azure service can be leveraged to analyse photo/image for extracting text, tags or object.
2. Custom Vision: Leverage this service to train personalized models for image classification and object detection using your own image datasets.
3. Face: This service enable to build solution for detecting face and facial recognition.
4. Form Recognizer: Utilize this service to extract text from scanned image or invoices.

Natural Language Processing

This is an area of AI which helps in creating software for understanding written or spoken language. It helps in creating software solution for:

1. Analyzing and interpret text in document, text messages or other data sources.
2. Interpret spoken language and synthesize speech response.
3. Automatically translate spoken or written language.
4. Interpret command and determine appropriate action.

Microsoft Azure Natural Language Processing Services:

1. Language: This service can be used to understand analyze text, training language model that can understand spoken or written text.
2. Translator: This service can be used to translate text in 60 different language.
3. Speech: This service helps to recognize and synthesize speech and translate spoken language.
4. Azure Bot: This service can be used to create Conversational AI based software agent which can participate in conversation.

Knowledge Mining

This is a process of extracting data and indexing it to make it searchable from large volume of unstructured data.

Microsoft Azure Knowledge Mining Services

1. Azure Cognitive Service: This service has tools for building indexes which can be used to create indexes which leads to make searching content easier.