| CODE | CCE3503 | ||||||||||||||||
| TITLE | Practical Machine Learning | ||||||||||||||||
| UM LEVEL | 03 - Years 2, 3, 4 in Modular Undergraduate Course | ||||||||||||||||
| MQF LEVEL | 6 | ||||||||||||||||
| ECTS CREDITS | 5 | ||||||||||||||||
| DEPARTMENT | Communications and Computer Engineering | ||||||||||||||||
| DESCRIPTION | This unit covers common methods and techniques used during the process of developing a pattern recognition data-driven model, including the design and of evaluation experiments. Topics include dataset analysis, scaling and feature selection, and statistical methods of analyse output. This unit does not cover the pattern recognition itself, but uses ready made implementations of the models. It is therefore a companion and follow-up unit to CCE2502 or equivalent (e.g CCE3502) in undergraduate foundations in Pattern Recognition and Machine Learning Models. Study-unit Aims: - To teach how to develop in a principled way, a data-driven pattern recognition model covering all important steps, from data to model output; - To provide students with an opportunity to practice the methods learned in a real-world problem setting via a practical assignment. Learning Outcomes: 1. Knowledge & Understanding By the end of the study-unit the student will be able to: - Explain covariance and correlation between features; - Describe how to select the important features; - Describe how to project features in a new n-dimensional space; - Describe when data normalisation is needed; - Describe when binning is required during data preparation; - Discuss the importance of an appropriate sampling process in dataset preparation; - Describe the distribution of a dataset; - Differentiate between uniform and stratified sampling; - Describe why a hold-out test set is necessary; - Describe why the dataset is split into train/val/test portions; - Describe why k-fold development is desirable for small datasets; - Describe how to handle categorical features in continuous input models; - Explain the appropriate performance measures for categorical and continuous targets and how to select the appropriate ones; - Explain the difference between single and multi class problems and multi-label problems. 2. Skills By the end of the study-unit the student will be able to: - Visualize relationships between features; - Compute Covariance and correlation between features; - Perform normalisation, binning and or sampling to data; - Compute the dataset distribution; - Shuffle and Split dataset into train/val/test portions; - Optimise hyperparameters in k-fold loop; - Design an evaluation experiment, including choosing the appropriate metrics; - Setup experiments to compare models and human evaluations and choose appropriate statistical metrics; - Evaluate models after deployment; - Use python libraries and tools such as sckit-learn, numpy, matplotlib; - Programming in Python or equivalent. Main Text/s and any supplementary readings: Main Text: - J.D. Kelleher, B. Mac Namee and A. D’Arcy, “Fundamentals of Machine Learning for Predictive Data Analytics”, MIT Press. (main library – Q325.5 .K455) |
||||||||||||||||
| ADDITIONAL NOTES | Pre-Requisite Study-units: CCE2502 or CCE3502 | ||||||||||||||||
| STUDY-UNIT TYPE | Lecture and Independent Study | ||||||||||||||||
| METHOD OF ASSESSMENT |
|
||||||||||||||||
| LECTURER/S | Trevor Spiteri |
||||||||||||||||
|
The University makes every effort to ensure that the published Courses Plans, Programmes of Study and Study-Unit information are complete and up-to-date at the time of publication. The University reserves the right to make changes in case errors are detected after publication.
The availability of optional units may be subject to timetabling constraints. Units not attracting a sufficient number of registrations may be withdrawn without notice. It should be noted that all the information in the description above applies to study-units available during the academic year 2023/4. It may be subject to change in subsequent years. |
|||||||||||||||||