Aaron Schumacher Builds TensorFlow Systems from Components at O'Reilly's OSCON
On May 9th, 2017, at O'Reilly's OSCON, Aaron Schumacher takes a building-block approach to exploring the tools TensorFlow provides so you can build the systems you need and write your own TensorFlow—not just run other people's scripts. Aaron discusses the many aspects of TensorFlow—including data management, machine learning, distribution, and serving—by comparing them with similar functionality in other toolkits.
Then on May 10th, Aaron talks more about product development at start-up Deep Learning Analytics, which has built a range of deep learning products since its founding in 2013. The Deep Learning Analytics team has built products with toolkits ranging from cuda-convnet to TensorFlow. Systems built on Caffe have matured and provide points of reference for comparison. Aaron Schumacher explains why TensorFlow is being chosen for more projects based on design strengths and features that will support future growth.