We introduce the computational network (CN), a generalization of popular machine
learning models such as deep neural network, recurrent neural network, convolutional
neural network, and log linear model that can be expressed as a series of computation
steps. We describe the benefits of such generalization and the key operations
on the CN.
We further introduce the computational network toolkit (CNTK), a general purpose C++ implementation of computational networks. We describe its architecture and core functionalities and demonstrate that it can construct and learn models of arbitrary topology, connectivity, and recurrence. The toolkit will be released under a modified Microsoft Research license agreement for non-commercial use.
Bibliographic reference. Yu, Dong / Eversole, Adam / Seltzer, Michael L. / Yao, Kaisheng / Guenter, Brian / Kuchaiev, Oleksii / Seide, Frank / Wang, Huaming / Droppo, Jasha / Huang, Zhiheng / Zweig, Geoff / Rossbach, Chris / Currey, Jon (2014): "An introduction to computational networks and the computational network toolkit (invited talk)", In INTERSPEECH-2014 (abstract)