The Tensor Deep Stacking Network Toolkit

In this paper we introduce the Tensor Deep Stacking Network (T-DSN) Toolkit, an implementation of the T-DSN deep learning architecture. The toolkit consists of a Python library and a set of accompanying helper scripts that allow you to train and evaluate T-DSN models. The toolkit is designed to be portable, modular, efficient and parallelized. Our goal for the toolkit is to promote research on this and related deep learning architectures. The T-DSN Toolkit is open source and free for non-commercial use. In this paper, we summarize the core functionality of the toolkit and discuss its design and implementation. We also present a new set of experiments on standard machine learning datasets, demonstrating the model´s effectiveness.