Toward a de facto standard in AI: What’s new in TensorFlow 2.0 (sponsored by IBM)

Toward the end of 2015, Google released TensorFlow, which started out as just another numerical library, but has grown to become a de facto standard in AI technologies. TensorFlow’s initial release was widely hyped, in no small part because it was released by Google, but the project also received complaints on usability—particularly, the fact that debugging was only possible after construction of the static execution graph.

In addition, neural networks needed to be expressed as a set of linear algebra operations, which was considered too low level by many practitioners. PyTorch and Keras addressed many of the flaws in TensorFlow and gained a lot of ground. The recent TensorFlow 2.0 has successfully addressed those complaints and promises to become the go-to framework for many AI problems.

Romeo Kienzler introduces you to the most prominent changes in TensorFlow 2.0 and explains how you can use these new features successfully in your projects, covering eager execution, parallelization strategies, the advantages of the tight high-level Keras integration, live neural network training monitoring using TensorBoard, automated hyperparameter optimization, model serving with TensorFlow service, TensorFlow.js, and TensorFlow Lite. He then shares an outlook on TFX—where Google is planning to open source its complete AI pipeline—and contrasts it with existing de facto standard frameworks like Apache Spark.

This session is sponsored by IBM.