University of Dayton – Machine Learning for Pattern Classification 2017-12-07T11:40:06+00:00

Project Description

Machine Learning for Pattern Classification

Contact us if no time left
Time left to submit challenges
0
0
0
0
Days
0
0
Hrs
0
0
Min
0
0
Sec

Summary

This course introduces the fundamental concepts and models of machine learning with a practical treatment of design, analysis, implementation and applications of algorithms which learn from examples. Topics include supervised and unsupervised learning, self organization, pattern association, feed-forward and recurrent architectures, manifold learning, dimensionality reduction, and model selection.

Student Insights

# StudentsDegreeField of StudyCollaboration Period (days)International Students (%)
25MasterComputer Science, Engineering6025

Description

1. Basic concepts of machine learning, machine perception, principles of learning and adaptation, supervised and unsupervised learning.
2. Linear discriminant functions and decision surfaces, generalized linear discriminant functions, minimization of perceptron criterion function, gradient descent algorithm.
3. Multilayer neural networks, feed-forward operations and classifications, back-propagation learning algorithm, back-propagation as feature mapping, radial basis function networks.
4. Recurrent neural networks, Hopfield network, Hebb’s learning rule, stability criterion, pattern association, bidirectional associative memory.
5. Adaptive resonance theory, vigilance criteria, self organization, fuzzy ART and ARTMAP, ARTMAP as supervised learning network, Kohonen’s self-organizing feature map.
6. Manifold learning, manifolds of perception, linear and nonlinear manifolds, local embedding, recurrent architectures for nonlinear manifolds.
7. Dimensionality reduction methods, component analysis and discriminants, principal component analysis, linear discriminant analysis, expectation minimization, hidden Markov models.

Objectives

This course requires a lot of interaction. Sharing of ideas is encouraged. This is a project oriented course. There will be 10 projects including the final project. All the projects should be implemented in MATLAB or C/C++.

Expected solution proposals

Project Report: The methodology, program outline with flow chart and/or illustrations, implementation results with sample data sets, comments / discussions on the obtained results, and appropriate technical references should be included in the project report and submitted as hardcopy submissions

Methods

MATLAB or C/C++.

Top Competencies

Critical thinking / problem-solvingCuriosityJudgement and Decision Making