Lecture 1: Introduction to Machine Learning

Introduction to the Module

  • Lecturer: Dr. Jing Wang (Module Leader) and other instructors (Dr. Ilya Alexakhin, Dr. Hooman Oroojeni Mohama Javad, Dr. Ala Barzinji).
  • Module Structure:
    • Lectures and IT labs are in person.
    • 12 Weekly Lectures on Monday in KW303 at 09:00-11:00.
    • Tutorials in the IT Labs.
    • Access to Moodle page COMP-1816-M01-2024-25.
    • Announcements via Moodle.
  • Module Contacts:
    • Use Greenwich email account to send email.
    • Allow instructors 3 working days for a reply.
    • Student support, IT problems, and administrative problems should be directed to specific contacts.
  • Schedule:
    • Weekly topics including Introduction, Supervised Learning, Regression, Multivariable Models, Model Selection, Classification, Evaluation metrics, Optimisation, Overfitting, Neural Network, Module Assessment Q&A, Advanced Neural Networks, Unsupervised Learning, Clustering, Dimensionality reduction, and Guest lecture.
  • Assessment:
    • Coursework (60%) and Exam (40%).
    • Coursework released on Monday 20/01/2025, submission deadline: Monday 10/03/2025 at 23:30 (UK Time).
    • Late submissions are capped at 40%.
    • Exam: Close-book, location: Greenwich Maritime Campus, date: TBC in April/May (UK time).
    • Resits available with deadlines in July.
  • Ground Rules:
    • Engagement is crucial.
    • Follow instructions on Moodle.
    • Recorded lectures and materials online.
    • Be considerate.
    • Avoid plagiarism.
  • Using ChatGPT Sensibly:
    • Use appropriately for understanding content, debugging code, and improving solutions.
    • Don't copy verbatim or assume it's always correct.
  • How to Succeed:
    • Engage with the Module by attending lectures and lab tutorials.
    • Arrive on time, pay attention, take notes, and ask questions.
    • Complete tutorial tasks and engage with tutors. Start coursework early, read the specification carefully, and ask questions.
    • Understand lecture contents, especially theories underlying algorithms.
  • Registering Attendance:
    • Register attendance during lectures/lab tutorials using the card reader or QR code.

Introduction to Machine Learning

  • What is Machine Learning?
    • Alan Turing Quote: "Instead of trying to produce a programme to reproduce the adult mind, why not rather try to produce one which simulates the child’s?"
    • Machine Learning is learning the rules plus experience.
  • Example: House price Prediction
    • Task (T): Prediction.
    • Experience (E): Known information/data.
    • Performance measure (P): Function evaluating performance on task T.
  • Technical Definition:
    • A computer program learns from experience E with respect to tasks T and performance measure P if its performance at tasks in T, as measured by P, improves with experience E (Mitchell, 1997).
  • Why Shall the Machine Learn?
    • To address cases where humans can't find or explain the observed data.
    • When humans can’t explicitly encode their expertise in computer algorithms.
    • To analyze vast volumes of data and produce fast and accurate results.
  • Applications of Machine Learning
    • Speech Recognition, Natural Language Processing, Computer Vision, Robotics, Medical Outcomes’ Analysis, Computational Biology, Sensor Networks, etc.
  • Case Studies:
    • Speech recognition, Machine Translation, Handwritten digit recognition, Face recognition, Face detection, Objective detection, Pose Estimation, Robotics.
  • Achievements by Machine Learning (Industry):
    • Various milestones from 1959 to 2018, including the development of chatbots, defeating chess champions, image recognition, and medical applications.
  • Achievements by Machine Learning (Academia)
    • Artificial intelligence for social good.
  • Connections to Other Subjects:
    • The module contains mathematical and statistical content, and help is available at mash@greenwich.ac.uk.
  • Work Still to Be Done:
    • Face recognition and image understanding are not always perfect.

Machine Learning Techniques

  • Overview of Machine Learning:
    • Meaningful Compression, Structure Discovery
  • Types of Machine Learning:
    • Supervised Learning: Learn through data feature and Target. Perform Task on new data. (Regression and Classification)
    • Unsupervised Learning: Data has no target. Learn based on the features only. Perform Task on all data. (Clustering and Dimensionality reduction)
  • Supervised Learning
    • Learn through data feature and Target.
    • Perform Task on new data.
    • Regression and Classification.
  • Unsupervised Learning
    • Data has no target, learn based on the features only.
    • Perform Task on all data.
    • Clustering and Dimensionality reduction.

Summary

  • Module Assessment: Coursework (60%) + Exam (40%)
  • Coursework submission deadline: Monday 10/03/2025 at 23:30 (UK time)
  • Exam date: A date in April/May (UK time)
  • Machine Learning requires E (experience/data), T (Task), and P (Performance evaluation).
  • Supervised Learning: E: data (features and target), T: on new data, P, evaluation metrics (classification, regression)
  • Unsupervised Learning: E: data (features only), T: on all data, P: evaluation metrics (dimensionality reduction, clustering)