Introducing TensorFlow

TensorFlow is a library for deep learning and was first released by Google in 2015. Initially, it included a core library that allowed users to work with tensors (that is, multidimensional arrays) in symbolic form, thus enabling low-level neural network design and training at high performance. Nowadays, it's a fully fledged deep learning library that allows data scientists to build models for complex problems, such as image recognition, using high-level primitives. You can also use TensorFlow for solving standard machine learning problems, such as the ones we've considered in previous chapters. TensorFlow has similar abstractions to the ones we have been using in scikit-learn, Apache Spark, and SageMaker. For example, it allows the user to create classification models or regression models using high-level abstractions, such as estimators, predictors...