cst 383: learning log 7
This week, was mainly focused on finals prep and also studying for the quiz. The focus was Logistic Regression. It is mainly used for predicting categories. This is a good way to predict inside of a standard 0 -> 1 range, since linear models don't do this. The Sigmoid function can dissect the output in the 0 -> 1 range. Additionally, we also covered underfitting and overfitting, and looking at learning curves which tests errors against various sizes of datasets. I learned that if there is a huge gap in between the two curves, there's an issue with variance, and we may need more data or just a simpler model. And if both of the errors are high and near each other, there could be a need for a more flexible model. It really does depend on the outputs. Finally I learned about translating data into a format that is actually processable. Personally I also did some research on the difference between accuracy and roc auc, because I learned that accuracy just focuses on a percentag...