/CSE-6363_MACHINE-LEARNING

Primary LanguageJupyter NotebookMIT LicenseMIT

CSE-6363_MACHINE-LEARNING

Project 1

  1. Write a compute program to solve theequations of Exercise 1.1, for the 10 data points of Figure 1.2 (you generate them according to the description there) for M=0,1,3, 9. Plot the fit curves and original data points as Figure 1.4.
  2. Modify the compute program of HW2 to solve the equations of Exercise 1.2, forthe 10 data points you generated in part 1A. Here the main focus is theregularization M=9. Show that as lambda of Eq.(1.4) increases, the overfit of Fig 1.4 (right-bottom figure) reduced significantly, to something like Figure 1.7.

Project 2

Write a computer program that can read in data such as those in Table 4.1. (A) The program can represent each data instance ---these are “stored data”(B) When a new data instance is presented, the program compares the new data instance to every “stored data instances”to compute the distance, and find out the k nearest neighbors, and predict the class label for the new data instance. (C) Compute the multinomialdistribution for each attribute of the data instances and prior probability. When a new data instance is presented, compute the class label using NAÏVEBayes classification method.

Project 3

Data Cluster using K-means algorithm provided by the system.1. Run k-means on AT&T 100 images, set K=10. Obtain confusion matrix. Re-order the confusion matrix using bipartite graph matching and obtain accuracy.2. Run k-means on AT&T 400 images, set K=40. Obtain confusion matrix. Re-order the confusion matrix and obtain accuracy.3. Run k-means on Hand-written-letters data, set K=26, as above.Computer Exam3 will depend on the codes you write for Project 3.