A Survey of Repair Strategies for Deep Neural Networks
Over the past decades, there has been a rapid increase in the development of neural network architecture and algorithms allowing its widespread use in different application domains. In the recent years, neural networks have seen their application in safety-critical systems such as autonomous driving systems, unmanned aerial vehicle collision avoidance, and facial recognition based on authentication systems. In order to ensure the correct behavior of such systems, it is, therefore, necessary to provide assurance of correctness of neural networks that are being deployed by these systems. In other words, if a neural network does not produce the expected result, it is necessary to develop strategies to "fix" the network effectively and efficiently. In this report, we review and critique different types of neural network repair strategies with the objective to present a comparative understanding of the capabilities of the strategies and the challenges that are yet to be resolved.
Committee: Samik Basu (major professor) and Guang Song
Join on Zoom: Please click this URL to start or join. https://iastate.zoom.us/j/98898833268 Or, go to https://iastate.zoom.us/join and enter meeting ID: 988 9883 3268