| (One intermediate revision by the same user not shown) | |||
| Line 25: | Line 25: | ||
*DO NOT PLAGIARIZE!!!!! You will get in serious trouble if you do so. Don't use figures from the internet, don't copy and paste other people's text, cite all your sources, etc. In doubt, ask your instructor. | *DO NOT PLAGIARIZE!!!!! You will get in serious trouble if you do so. Don't use figures from the internet, don't copy and paste other people's text, cite all your sources, etc. In doubt, ask your instructor. | ||
| − | *Joe will be sharing some MATLAB code to compute Bayes Error in multiple dimensions. (Will be posted here shortly.) Feel free to use his code (with citation, of course!) | + | *Joe will be sharing some MATLAB code to compute Bayes Error in multiple dimensions. (Will be posted [[bayes_error_MATLAB_code_joe|here]] shortly.) Feel free to use his code (with citation, of course!) |
*In the past, I have been asked if it is ok to use Naive Bayes for the classification. You can use it if you are trying to compare this approach with the standard Bayes Decision Rule, but do not do the entire project with Naive Bayes, it is too simplistic. | *In the past, I have been asked if it is ok to use Naive Bayes for the classification. You can use it if you are trying to compare this approach with the standard Bayes Decision Rule, but do not do the entire project with Naive Bayes, it is too simplistic. | ||
| + | |||
| + | *If you have no idea how to generate data samples from two classes (esp. how to deal with the priors), take a look at this [[Generating_random_data_with_controlled_prior_probabilities_slecture_ECE662S14_Gheith|slecture by Alex Gheith]]. For example, if the priors are equal (50% each), you should '''not''' always have the same number of points for each class! | ||
| + | |||
---- | ---- | ||
[[2016_Spring_ECE_662_Boutin|Back to ECE 662 Spring 2016]] | [[2016_Spring_ECE_662_Boutin|Back to ECE 662 Spring 2016]] | ||
Latest revision as of 16:19, 1 February 2016
First Mini-Project, ECE662 Spring 2016
Due in class Friday February 12. Earlier submissions are welcome!
Note about late submissions: Late submissions will be accepted until 5pm Thursday February 18 in MSEE330. This is a hard deadline. NO EXCEPTION! Please try not to be late and hand in by the February 12 deadline.
Question
We have learned how to use Bayes Decision Rule to classify data points drawn at random from two classes that are normally distributed in $ R^n $. Use data (either synthetic or real) to experiment with this classification method. When does it work well? When does it not work well?
Write a report to explain your experiments and summarize your findings. Make sure to include a cover page, introduction, numerical results (e.g., tables, graphs,…), and a conclusion. Attach a copy of your code.
Hand in two hard copies of your report. The first one should start with a standard cover page. For the second one, remove your name from the cover page. Place the anonymous submission below the submission with a name and hand in both together.
- Specific examples of project scopes and experiments will be described in class. Feel free to come up with your own.
- You must write your own function to classify the data (discriminant g(x)). Do not copy other people's code and do not use any toolbox to classify the data.
- DO NOT PLAGIARIZE!!!!! You will get in serious trouble if you do so. Don't use figures from the internet, don't copy and paste other people's text, cite all your sources, etc. In doubt, ask your instructor.
- Joe will be sharing some MATLAB code to compute Bayes Error in multiple dimensions. (Will be posted here shortly.) Feel free to use his code (with citation, of course!)
- In the past, I have been asked if it is ok to use Naive Bayes for the classification. You can use it if you are trying to compare this approach with the standard Bayes Decision Rule, but do not do the entire project with Naive Bayes, it is too simplistic.
- If you have no idea how to generate data samples from two classes (esp. how to deal with the priors), take a look at this slecture by Alex Gheith. For example, if the priors are equal (50% each), you should not always have the same number of points for each class!
