DARVIZ: Deep Abstract Representation, Visualization, and Verification of Deep Learning Models

TL;DR

DARVIZ introduces a comprehensive framework for visualizing, interpreting, and verifying deep learning models to address the challenges of understanding complex data-driven software systems across multiple libraries.

Contribution

It presents a novel approach for unified visualization and verification of deep learning models across diverse frameworks, enhancing interpretability and interoperability.

Findings

Improved understanding of deep learning model behavior

Enhanced interoperability across deep learning libraries

Facilitated debugging and verification of models

Abstract

Traditional software engineering programming paradigms are mostly object or procedure oriented, driven by deterministic algorithms. With the advent of deep learning and cognitive sciences there is an emerging trend for data-driven programming, creating a shift in the programming paradigm among the software engineering communities. Visualizing and interpreting the execution of a current large scale data-driven software development is challenging. Further, for deep learning development there are many libraries in multiple programming languages such as TensorFlow (Python), CAFFE (C++), Theano (Python), Torch (Lua), and Deeplearning4j (Java), driving a huge need for interoperability across libraries.