Assignment for machine learning course, Robotic Engineering second year.
This assignment is based on naive Bayes classifier that is a probabilistic machine learning algorithm based on Bayes Theorem. It is called ”naive” because it operates on the assumption of independence of the features that is not always true. The requirement was to implement its algorithm as a Matlab function. We had also to implement in a second function an improved version of the classifier that tries to overcome some of its limitations. The final results show that the first function works well in normal situations, but when we have in input a particular data set the improved version works better. *
This assignment is based on linear regression algorithm. There are different type of algo- rithm in order to adapt to different cases. The requirement of this assignment was to discover these algorithms, one-dimensional problem with or without intercept, multi-dimensional prob- lem on a full multi-column dataset. At the end, it was required also to discover the average of the Mean Square Error of a part of the database (10% and 90%).
This assignment is based on the k-Nearest Neighbor classification algorithm with the MNIST dataset of the handwritten digits, that it's a non-parametric method used for classifi- cation when the complexity depends on the numbers of available data. We used the Euclidean distance metric and we, also, use a target. At the end, we performs test in order to find which k value was the best and other feature, however, the biggest problem was computational time!