A broad introduction to machine learning and statistical pattern recognition. Topics include supervised learning (the use of labeled datasets to train algorithms to classify data or predict), unsupervised learning (uses machine learning algorithms to analyze and cluster unlabeled datasets), learning theory (bias/variance tradeoffs), and practical advice.
Additional Requirements for Graduate Students: Graduate students are required to propose their own pattern recognition engineering project based on a literature survey of selected course topics. They are expected to design and develop their own code, share it for assessment, present their work, and submit a final project report.
Athena Title
Pattern Recognition
Prerequisite
(CSCI 1301-1301L or CSCI 1301E or ELEE 2040 or INFO 2000 or INFO 2000E) and ENGR 2090
Semester Course Offered
Offered every year.
Grading System
A - F (Traditional)
Student Learning Outcomes
Ability to select and implement pattern recognition techniques in a computing environment, which is suitable for engineering applications.
Ability to prototype, model, and test various pattern recognition algorithms.
Ability to identify the characteristics of datasets, pre-process them, and compare the impact of using different datasets on pattern recognition system performance.
Ability to integrate pattern recognition libraries and mathematical and statistical tools with modern technologies (MATLAB, Python).
Ability to understand different types of metrics available to evaluate the performance of pattern recognition engineering solutions.
Ability to select a prototyped model and make it work on a practical scenario, first at a smaller, and then at a larger, scale.
Additional course objectives or expected learning outcomes for Graduate Students:
Ability to solve problems associated with batch learning and large-scale data characteristics, including high dimensionality, dynamically growing data, and scalability issues.
Ability to understand and apply scaling up pattern recognition approaches and associated computing techniques and technologies (e.g., data augmentation).
Ability to recognize and implement various ways of selecting suitable model parameters for different machine learning techniques.
Graduate students will also perform a relevant literature study and project related to the course topics listed.
Topical Outline
This course will introduce a graduate audience to salient topics in Machine Learning and Pattern Recognition (Supervised and Unsupervised Learning):
Supportive Material to Linear Algebra and Probability Theory
Introduction to Pattern Recognition
Linear Regression
Logistic Regression
Gradient Descent
Neural Networks
Support Vector Machines (SVMs)
Bayesian Decision Theory
Linear Discriminant Functions
Clustering
Dimensionality Reduction
Principal Component Analysis and Multidimensional Scaling
Density Estimation Schemes
Nearest-Neighbor Rule
Feature Extraction
Pattern Recognition Case Studies (Invited Speakers/Experts)
The topics will be taught not necessarily in the above order.
The project component of this course will test the student's ability to design and evaluate classifiers on appropriate datasets.
