Activity: Talk or presentation types › Invited academic talk
Description
This lecture decomposes machine learning (ML) into three main components : data, model and loss. Every ML method uses specific design choices for data (representation), model and loss function. A main design principle for ML methods is empirical risk minimization (ERM). ERM learns optimal model parameters by minimizing the average loss on a training set which consists of carefully chosen data points.
|Lecture could not be delivered in person due to a sick leave.]