Machine Learning in Elixir

Chapter 11Model Everything with Transformers

At this point in the book, you’ve worked with many different types of data, such as structured data, visual data, text data, and time-series data. For each of these specific applications of machine learning, you used a new approach or architecture to solve the problem. With each of these new models or architectures, you saw firsthand how building a model with some simple assumptions and slight tweaks can drastically impact its performance.

In Chapter 7, ​Learn to See​, you used convolutions with learned kernels to drastically increase the ability of your model to extract features from image inputs. Convolutional neural networks work well because they assume your inputs have a grid-like structure. ...