Intro to Machine Learning

Introduction to Machine Learning

• This is an introduction to the field of machine learning
• Aims to provide a unified view of the field and to present key ideas and techniques
• Will be example driven but also cover theoretical concepts
• The module content overview, What and Why of machine learning, and ingredients of machine learning (tasks, features, and models) will be covered

Mid-Module Evaluation

• Module is about halfway through
• There is an opportunity to provide feedback on how the module is going
• Brief mid-module evaluation survey on Canvas (or a direct link)
• Survey is anonymous and helpful to improve the course
• Survey open and closes next Sunday at 4 PM

Data Science Process

• The data science process is cyclical
• Real world data, collection, wrangling, and cleaning then exploratory data analysis, statistical inference and hypothesis testing
• The cycle continues with mathematical modelling and data solutions/products.

What is Machine Learning?

• A subset of AI involving computer programs (models)
• Models automatically learn from data and perform tasks requiring few explicit instructions
• Learns patterns and performs recognition
• Learns a generalized mechanism to solve a task

Why is Machine Learning Interesting?

• Potential to automate trivial, repetitive and dangerous tasks
• Fast processing of information
• Vast amount of data being generated and stored digitally
• Opportunity for deeper insights from more data

Why Should Machine Learning Be Interesting To You?

• Curiosity and designing frameworks to learn from data
• Ubiquitous use of machine learning systems
• Combination of programming, applied mathematics and machine learning skill sets
• Good job prospects in the field

How do you interact with ML in your Life?

• A poll is used find out about machine learning interactions in daily life
• Helps understand different machine learning tasks and requirements

What applications of ML have you used?

• A poll to find out what ML applications users have used

Outline for Today

• The rest of the module content overview
• The What and Why of machine learning
• Ingredients of machine learning (tasks, features, and models)

Learning Outcomes

• Understand the variety of tasks in machine learning (classification, regression, clustering and dimensionality reduction)
• Learn what machine learning models are
• Recognize the importance of features (dimensions) in machine learning

About This Section of the Module

• This is an introduction to machine learning
• An example-driven approach, but will also cover important theoretical concepts

Learning Outcomes

• Basic knowledge of supervised and unsupervised machine learning models (linear regression, perceptron, random forest, PCA, and K-means clustering)
• Map machine learning models to tasks
• Explores concepts like cross-validation and learning curves
• How to use machine learning toolboxes
• Preprocessing and incorporating prior knowledge into solving classification/regression problems with real-world data

Use the Resources Provided

• Access to teaching materials, solutions, assessments, details and further reading on Canvas
• Module Discussion Forum to ask questions and share resources

Source Materials

• Recommended books and online resources for learning machine learning
• Includes titles and URLs

Mathematics and Machine Learning

• Machine learning is underpinned by mathematical theories and formalisms
• Essential mathematics (probability theory, linear algebra, multivariable calculus, information theory, and logic)

The Reminder of the Module

• Week-by-week outline of module content, including topics, lecturers, and supervised and unsupervised learning.

Assessment of a Noninvasive Exhaled Breath Test for the Diagnosis of Oesophagogastric Cancer

• Research article, discussing a noninvasive exhaled breath test for diagnosing oesophagogastric cancer

Supervised Learning

• Training paradigms
• Data has associated labels/classification
• Model generalizes on unseen data

Unsupervised Learning

• Machine learning does not require labels
• Learns patterns and structures from similarities and differences in training examples

Degrees of Supervision

• Semi-supervised machine learning
• Using labels on only a subset of data
• Relevant when labels are expensive.

Data and Features

• Features are the inputs to machine learning
• Used to describe data objects
• Numerical, boolean, ordinal, or nominal
• It's crucial to choose the right features

Machine Learning Models

• Models are central in machine learning
• A framework to predict from features using parameters learned from data
• Models encode the task mapping

Quiz Time! (Multiple Choice Question)

• Question about which machine learning task the weather forecast data is related to

Additional Important Tasks

• Classification
• Regression
• Clustering, Collaborative filtering
• Dimensionality reduction

What problems can ML solve that were previously considered very difficult?

• A question

Why Machine Learning?

• Data can be used to perform all sorts of tasks
• Identifying irrelevant information (spam)
• Predicting weather (sunny, partly cloudy, rainy, thunderstorm), temperature, wind, and humidity
• Recommending products or services (purchase history)
• Grouping images based on semantic similarity
• Visualizing high-dimensional data
• Solving problems that are difficult to formalize

Tasks, Models, and Features

• The essence of machine learning
• Tasks—mapping data to desired outputs
• Features—characteristics of data
• Models—required mapping/task encoding

(Some) Machine Learning Tasks if You have Health Data

• Tasks specific to health data
• Classification, binary classification, regression, clustering, collaborative filtering, dimensionality reduction, and reinforcement learning