TensorFlow and its implications to libraries

TensorFlow and its implications to libraries

TensorFlow, which stemmed from group research within Google, is a machine learning library made open source at November in the year of 2015. With TensorFlow, data flow chart can be constructed, and then executed after a session is launched. Meanwhile, data can also be fed, and fetched when a program is running. An edge represents a flow of tensor and a node represents a mathematical operation which takes and/or produces one or more tensors. The core idea of constructing a data flow chart make it less likely to be error-prone. Data charts can also be visualized in Tensorboard, a visualization shipped along with TensorFlow. This session is intended to be primarily explorative. It will provide introduction to TensorFlow, (demonstration will be made through Python, one of the Application Program Interfaces), how it can potentially benefit existing library systems, and lessons learned along the way. TensorBoard will be covered if time permits.

Jiang, Minghao

Chinese American Librarians Association

2016-05-13

https://creativecommons.org/licenses/by-sa/4.0/

pdf

eng

text

2016 CALA Midwest Chapter Annual Conference "Innovate! Advocate! Integrate! – Reinvigorating Your Library in a Digital Age"