Estimate and Replace: A Novel Approach to Integrating Deep Neural Networks with Existing Applications

TL;DR

This paper introduces 'Estimate and Replace', a method that integrates existing applications into deep neural networks by estimating their functionality with neural estimators during training, then replacing them with actual applications at inference for improved performance.

Contribution

The paper proposes a novel approach to incorporate existing applications into deep learning models through estimation and replacement, enabling better utilization of existing knowledge and improved efficiency.

Findings

Outperforms non-interacting DNNs in accuracy

Requires less training data

Successfully integrates external applications during inference

Abstract

Existing applications include a huge amount of knowledge that is out of reach for deep neural networks. This paper presents a novel approach for integrating calls to existing applications into deep learning architectures. Using this approach, we estimate each application's functionality with an estimator, which is implemented as a deep neural network (DNN). The estimator is then embedded into a base network that we direct into complying with the application's interface during an end-to-end optimization process. At inference time, we replace each estimator with its existing application counterpart and let the base network solve the task by interacting with the existing application. Using this 'Estimate and Replace' method, we were able to train a DNN end-to-end with less data and outperformed a matching DNN that did not interact with the external application.