Dive into Deep Learning - Lecture 4: Logistic/Softmax regression and Cross Entropy Loss with PyTorch
Just as we implemented linear regression from scratch, we believe that logistic regression and softmax regression are similarly fundamental. These classifiers are used for binary and multiclass classification problems in machine learning. We start by talking about the logistic or sigmoid function and then explain the predicted probabilities. We also explain the cross entropy loss that is widely used for classification problems. We then present a data science problem in which we work with the Fashion MNIST data set and define a simple neural network using torch.nn. We also define loss functions and optimizers to solve the problem. We use torch.autograd for finding partial derivatives and making predictions.
#LogisticRegression #SoftmaxRegression #CrossEntropyLoss