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

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.

References in corpus (7)

• Sequence to Sequence Learning with Neural Networks
• Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
• Categorical Reparameterization with Gumbel-Softmax
• Achieving Human Parity in Conversational Speech Recognition
• Neural Programmer-Interpreters
• Compositional Semantic Parsing on Semi-Structured Tables
• Making Neural Programming Architectures Generalize via Recursion

Cited by in corpus (11)

• Towards out of distribution generalization for problems in mechanics
• Max-cut Clustering Utilizing Warm-Start QAOA and IBM Runtime
• Numerical Solution and Bifurcation Analysis of Nonlinear Partial Differential Equations with Extreme Learning Machines
• The Demise of Single-Authored Publications in Computer Science: A Citation Network Analysis
• Embedded methods for feature selection in neural networks
• Forecasting in multivariate irregularly sampled time series with missing values
• Noise Optimization for Artificial Neural Networks
• Deep Learning Based Android Malware Detection Framework
• Distributional Clustering: A distribution-preserving clustering method
• ProblemChild: Discovering Anomalous Patterns based on Parent-Child Process Relationships
• Alternative Paradigms of Computation