5.1.2 Machine Learning

This topic explains Machine Learning (ML), a field within Artificial Intelligence that teaches computers to learn from data and experience without direct programming.

Definition

Machine Learning (ML): Machine learning is a field within Artificial Intelligence that teaches computers to learn from data inputs and experience like humans without direct programming. AI algorithms are based on ML to predict output.

Core Concept

Machine Learning is like teaching a child to recognize animals. You don't give the child a rulebook saying "a cat has pointy ears, whiskers, and a tail." Instead, you show the child many pictures of cats and say "this is a cat." Eventually, the child learns to recognize cats on their own. Similarly, ML algorithms are shown many examples (data) and learn to recognize patterns, make predictions, and improve over time. The more examples they see, the better they get—just like a child learning from experience.

Detailed Explanation

What is Machine Learning?

Machine learning is a field within Artificial Intelligence that teaches computers to learn from data inputs and experience like humans without direct programming. AI algorithms are based on ML to predict output.

• Field within AI.
• Computers learn from data and experience.
• No direct programming needed.
• AI algorithms based on ML to predict output.
• Improves over time with more data.

How Machine Learning Works

Machine learning algorithms facilitate computers to train on data inputs and analyze it to produce accurate output values. It is the technology that recognizes patterns in large databases to build models to automate decision-making. Machine learning algorithms improve their decision-making capabilities over time as more sample data is input in the computers.

• Trains on data inputs.
• Analyzes data to produce output.
• Recognizes patterns in large databases.
• Builds models to automate decision-making.
• Improves with more sample data.

Applications of Machine Learning

Machine learning has become a very useful technique for solving problems related to many areas. These include:

• Recognizing images and speech
• Detecting fraud in banking transactions
• Identifying spam and email phishing
• Image and video recognition
• Translation
• Virtual personal assistants
• Google search engine
• Self-driving cars
• Manufacturing
• Medical devices

Types of Machine Learning

There are two types of machine learning:

• Supervised Machine Learning
• Unsupervised Machine Learning

Supervised Machine Learning

Supervised learning algorithms take a set of known input data and known output responses to build a model to make predictions. In supervised learning, the goal is to learn how to do the mapping to match the input to the output. Once the model is built, it is used to make new predictions on unseen data as well. Machine learning is used to solve complex problems that involve large amounts of data but there is no formula to predict the output.

• Uses labeled data (known input and output).
• Learns mapping from input to output.
• Used for prediction on new data.
• Solves complex problems with large amounts of data.
• No formula to predict output.

Example: Predicting Car Prices

• Engine type and power
• Transmission type (manual/automatic)
• Number of seats
• Front-wheel/rear-wheel drive
• Keyless entry
• Push button start
• Safety features
• Country of manufacturing

We have to label the cars with their prices. After entering the data for a large number of cars, we can train the algorithm and create a model. The price of a new car can be predicted based on the input fed in the supervised learning algorithm.

Unsupervised Machine Learning

In unsupervised learning, the training data is unseen and it is un-labeled. The meaning of un-labeled is that the data is not assigned a category or group. The unseen data is fed into the machine learning algorithm to train the model. The trained model tries to search for a pattern and gives the desired result. The unsupervised learning algorithm looks for hidden patterns in the input data. It uses a learning technique known as clustering (grouping). The unsupervised learning algorithm finds hidden patterns or groupings in the input data to make them suitable for clustering and data analysis.

• Uses unlabeled data (no known output).
• Finds hidden patterns or groupings.
• Uses clustering technique.
• No predefined categories.
• Data is not assigned a category or group.

Example: Loan Repayment Prediction

A bank wants to predict how capable an applicant is of repaying the loan. A machine learning system is to be set up to ensure that the loan is given to the applicant who can repay it. In this case, the applicant will provide data about monthly income, debt, credit history, etc., along with personal data such as age, mobile number, gender, etc. Information about age, mobile number, and gender are not important for determining the loan risk. Therefore, age, mobile number, and gender can be eliminated for setting up the machine learning system for predicting the loan risk.

Machine Learning Process

The machine learning process consists of 5 stages as shown below:

a) Collection of training data b) Creating algorithm c) Learning process d) Creating training model e) Predicting results

• Training Data
• Training the Machine
• Building a Model
• Predicting Result

Comparison of Supervised and Unsupervised Learning

Intermediate Level

• Supervised Learning: Uses labeled data (known input-output pairs) to train a model that can predict outputs for new inputs. Examples: car price prediction, spam detection.
• Unsupervised Learning: Uses unlabeled data to find hidden patterns or groupings (clustering). Examples: customer segmentation, loan risk grouping.

The ML process has five stages: data collection, algorithm creation, learning, model creation, and prediction. ML is used in image recognition, fraud detection, translation, self-driving cars, and more.

Important Points

• - ML: field within AI; learns from data without direct programming.
• - Supervised: labeled data, predicts output.
• - Unsupervised: unlabeled data, finds patterns (clustering).
• - ML process: data collection, algorithm, learning, model, prediction.
• - Applications: image/speech recognition, fraud detection, spam filtering, translation, self-driving cars.
• - Supervised example: car price prediction.
• - Unsupervised example: loan risk grouping.
• - ML improves with more data.

Remember

• - ML: Learns from data.
• - Supervised: Labeled data.
• - Unsupervised: Unlabeled data (clustering).
• - ML Process: Data → Algorithm → Learning → Model → Prediction.
• - Applications: Image, Speech, Fraud, Spam, Translation, Cars.
• - Supervised Example: Car price prediction.
• - Unsupervised Example: Loan risk grouping.
• - More Data = Better Model.

Common Mistakes

• - Confusing supervised and unsupervised: Supervised uses labeled data; unsupervised uses unlabeled.
• - Thinking ML is the same as AI: ML is a subset of AI.
• - Assuming ML always works perfectly: ML can make errors due to bias or insufficient data.
• - Forgetting the importance of data quality: Poor data leads to poor models.
• - Mixing up classification and clustering: Classification is supervised; clustering is unsupervised.

Quick Revision

Machine Learning (ML) is a field within Artificial Intelligence that teaches computers to learn from data and experience without direct programming. It recognizes patterns in large databases and builds models to automate decision-making. There are two main types: Supervised Learning (labeled data, predicts output) and Unsupervised Learning (unlabeled data, finds patterns via clustering). The ML process has five stages: data collection, algorithm creation, learning, model creation, and prediction. ML is used in image recognition, fraud detection, translation, self-driving cars, and more.

60-Second Revision

ML: Learns from data. Supervised: Labeled data (predict output). Unsupervised: Unlabeled data (clustering). ML Process: Data → Algorithm → Learning → Model → Prediction. Applications: Image, Speech, Fraud, Spam, Translation, Cars. Supervised Example: Car price prediction. Unsupervised Example: Loan risk grouping. Key Point: ML enables computers to learn from data without explicit programming; supervised uses labeled data, unsupervised finds patterns; more data leads to better models.