Deep Learning with PyTorch

Model architecture for different machine learning problems

The kind of problem we are solving will decide mostly what layers we will use, starting from a linear layer to Long Short-Term Memory (LSTM) for sequential data. Based on the type of the problem you are trying to solve, your last layer is determined. There are three problems that we generally solve using any machine learning or deep learning algorithms. Let's look at what the last layer would look like:

• For a regression problem, such as predicting the price of a t-shirt to sell, we would use the last layer as a linear layer with an output of one, which outputs a continuous value.
• For classifying a given image as t-shirt or shirt, you would use a sigmoid activation function, as it ...