AI Assignment Part I

Yesterday night was the only time I could start working on the AI assignment. Part of the Assignment was code a Naive Bayesian Classifier. A classifier basically tries to classify data into its correct classification based on previous knowledge (ie training data set). Our task is that we are given bitmaps (16x16 images with either black or white) and we are to classify the bitmaps into respectively digits (Each bitmap is an image of a digit).

Yesterday night I was struggling, trying to understand all the stuff I have missed while I was in Manila and when I was rushing homework assignment in the morning. Shafeeq was telling me that Naive Bayesian Classifier has something to do with Bayesian Networks, but I found out because it is assumed that the data are independent of each other, it is not really a Bayesian Network. With some help from Mun Thye, I managed to come up with my first Bayesian Classifier. Its accuracy was quite low (65%). I went to central library to do research to improve its accuracy. With some changes of the underlying model, its accuracy suddenly jumped to 88% and I'm really happy about it.