Scribd
Upload
10 views

EE236 Course Project Total: 10% of Grade

The document describes the course project assignment for EE236, which is worth 10% of the total grade. Students must complete an independent research project involving mathematical modeling and MATLAB implementation to solve a problem related to estimating the parameters of a mixture of Gaussians model using expectation maximization. The project requires formulating assumptions, analyzing parameter performance through variation, and writing a 3-page maximum report divided into abstract, introduction, problem description, solution strategy, experimental results and analysis, conclusions, and references sections. While discussion with others is permitted, independent work and own experimentation and reporting is expected. Violations will be strictly punished.

Uploaded by

niloymithun
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
10 views

EE236 Course Project Total: 10% of Grade

The document describes the course project assignment for EE236, which is worth 10% of the total grade. Students must complete an independent research project involving mathematical modeling and MATLAB implementation to solve a problem related to estimating the parameters of a mixture of Gaussians model using expectation maximization. The project requires formulating assumptions, analyzing parameter performance through variation, and writing a 3-page maximum report divided into abstract, introduction, problem description, solution strategy, experimental results and analysis, conclusions, and references sections. While discussion with others is permitted, independent work and own experimentation and reporting is expected. Violations will be strictly punished.

Uploaded by

niloymithun
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 1

EE236 Course Project

Total: 10% of grade

Below is the description for the project problem. These will require some independent research,
mathematical formulation, and software implementation in MATLAB. You might find the problems to be
slightly vague. That is not unexpected in real-life applications. As needed, create your own assumptions
and constraints, explain them clearly in your report, and provide the solution. You will be judged by the
rigor of your analysis, e.g., which parameters did you vary to analyze performance. Report format will
be 12 point, Times Roman, single spaced with a maximum of 3 pages, including figures and references.
The report should be divided into the following segments: Abstract, Introduction, Problem description,
Solution Strategy, Experimental Results and Analysis, Conclusions, and References. While you may
discuss with other students in your class, you are expected to do your own experiments and write your
own report. Any violations will be dealt with very severely.
1. This problem is related to estimating the parameters of a mixture of Gaussians (MoG) model using
expectation maximization (EM).
1. Research and read up on MoGs and EM. There are many papers, books, and online resources.
You can just search online and find them. I particularly like the description in Pattern Classification
by Duda, Hart and Stork (the book is in the library).
2. Implement the EM algorithm. You need to actually write the code, not use the MATLAB or
other existing software function. You can use in-built functions for parts within the main EM algorithm.
3. Generate data from an MoG model. Assume you know the number of components of the MoG.
Use the EM to estimate the parameters of the MoG (parameters of each Gaussian and their mixing
values).
4. Study how the iterations in Step 3 progress and comment on the convergence of the parameters.
5. Repeat the parameter estimation part without accurate knowledge of the number of components.
Study and report what you observe about the parameter estimation process.

You might also like