ECE 417 Homework and ExamsHomework
- MP1, Due 9/12: Metric Learning, Mahalanobis Distance
- MP2, Due 9/26: Principal Component Analysis (PCA), K-Nearest Neighbors (KNN)
- MP3, Due 10/10: Cepstrum and Mel-Frequency Cepstrum (MFCC)
- MP4, Due 10/24: Gaussian Mixture Model (GMM)
- MP5, Due 11/7: Hidden Markov Model
- MP6, Due 11/28: Image Object Detection
- MP7, Due 12/12: Video Animation
Submit your MPs to Compass
Homework Rubric
"Best 6.5/7" plan: your final MP grade is the sum of 6 highest MPs, plus 0.5 times the lowest one.
Team grading: You can submit one assignment for two or three students working together. In this case, all of you will receive the same grade for that assignment. You are free to change partners between one assignment and the next.
Each MP is graded on a 15-point scale, with the following points:
- Correct Submission Format (1 point): You receive this point automatically if you manage to submit your homework in the correct format. The correct format is defined as follows:
- Upload two separate files on Compass. One file should be called MPx_yyy.pdf, the other should be called MPx_yyy.zip, where x is the MP number, and yyy is your NetID. Note that yyy should be the NetID of the person uploading the MP on Compass. This holds true even if your group has more than one member. Example: If your NetID is thanks@illinois.edu, then for MP1 you would turn in: MP1_thanks.pdf and MP1_thanks.zip. (sample copies)
- The PDF document should be no more than four pages in length, and should be a PDF document containing the sections ``I. Introduction,'' ``II. Methods,'' ``III. Results,'' and ``IV. Discussion.''
- The ZIP file should be a directory containing your runnable matlab code and a "members.txt" file. On each line of the members file, you will list the name of a group member followed by her/his NetID. An example of members file is here. The members file is the only place where we'll know who you collaborated with. If you worked alone, you will still have to provide the members file.
- Finally, do not include the distributed train/test datasets in your ZIP file. Your code should be able to read the necessary data from a data directory supplied by the TA. More information on this is given below.
- Introduction (2 points): (1) Give the name of the feature vector being used for either image or audio features in this MP. Spend at least one sentence giving an intuitive reason why that feature vector is more useful than raw audio samples (or raw pixels). (2) Give the name of the learning algorithm being used in this MP (the distance metric, classifier, signal generator, etc). Spend at least one sentence describing what that algorithm learns from its training data, and how it does so.
- Methods (4 points): (1) Identify line numbers, in your code, that implement at least two of the critical transformations described in the introduction (some parts of the feature computation or machine learning). (2) Include, in your narrative, equations that show the same two computations, but in mathematical rather than matlab notation. Your equations should be written using LaTeX, Microsoft Equation Editor, or some other true equation editor; plaintext equations are not allowed. Your equation should not include any multi-character variable names; each variable name should be one character.
- Results (2 points): (1) Give your complete results, in a table or chart. (2) Your results should be correct.
- Discussion (1 point): Pick one pair of numbers from your results, and write down a hypothesis or explanation specifying a physical or mathematical origin for the difference in those two numbers (why one of the numbers is larger than the other).
- Code (6 points): Provide your software in the form of a zipped directory. The TA should be able to unpack your archive using unzip, start Matlab, cd into the directory, and type run(datadir). There should be a function in your directory called run.m that will read the distributed data files from directory datadir (a string variable containing the directory name), and that will then generate exactly the same plots and/or tables that were in your narrative report.
- Extra Credit: Bonus points are available on many of the MPs. Each correctly completed bonus section adds up to 1 percent to your cumulative course total. Some MPs have one bonus section, some have two, some have none.
Exams
Best 2.5/3 plan: your final exam grade is the sum of 2 highest exams, plus 0.5 times the lowest one.
What you can bring to an exam: you can bring one page of notes, handwritten, front and back. Bring pencils and erasers. No calculators are permitted.
Regrade requests: After reading the solutions, if you think there is a mistake in the grading of your exam, attach a separate piece of paper to the front page of your exam, and hand it back at the next lecture after you get your exam.
- Exam 1, 9/28
- Fall 2017: Exam, Solutions, Conflict Exam, Conflict Solutions,
- Spring 2016 Exam 1 and its Solutions
- Spring 2015 Exam 1 and its Solutions
- Spring 2014 Exam 1 Solutions
- Exam 2, 10/26
- Exam 3, During Finals Week