Deep learning has successfully achieved state-of-the-art performance in a variety of applications, such as image classification, object detection, speech recognition, and so on. But deep learning architectures often fail when forced to make predictions about data for which there is little supervised information available. As we know, mathematics is fundamental to all machine learning and deep learning models; we convey our data and objectives to machines using mathematical representations of the data. These representations can have many forms, especially if we want to learn complex tasks (for example, disease detection), or if we want our architecture to learn representations based on different objectives, for example, to calculate the similarity between two images, we can calculate both Euclidean distances and cosine similarity.
In this chapter, we will...
