Neural Network Libraries: A Deep Learning Framework Designed from Engineers' Perspectives

TL;DR

This paper introduces Neural Network Libraries, a deep learning framework focused on usability and compatibility, addressing the evolving complexity of the field with flexible design and efficient distributed computation.

Contribution

It presents a new deep learning framework designed from engineers' perspectives, emphasizing usability, compatibility, and performance validation.

Findings

Framework enhances flexibility in network design

Achieves faster computation in distributed settings

Demonstrates improved usability and compatibility

Abstract

While there exist a plethora of deep learning tools and frameworks, the fast-growing complexity of the field brings new demands and challenges, such as more flexible network design, speedy computation on distributed setting, and compatibility between different tools. In this paper, we introduce Neural Network Libraries (https://nnabla.org), a deep learning framework designed from engineer's perspective, with emphasis on usability and compatibility as its core design principles. We elaborate on each of our design principles and its merits, and validate our attempts via experiments.