Vicent Sanz Marco

Tinghui Ouyang, Vicent Sant Marco, Yoshinao Isobe et al.

As the major factors affecting the safety of deep learning models, corner cases and related detection are crucial in AI quality assurance for constructing safety- and security-critical systems. The generic corner case researches involve two interesting topics. One is to enhance DL models robustness to corner case data via the adjustment on parameters/structure. The other is to generate new corner cases for model retraining and improvement. However, the complex architecture and the huge amount of parameters make the robust adjustment of DL models not easy, meanwhile it is not possible to generate all real-world corner cases for DL training. Therefore, this paper proposes to a simple and novel study aiming at corner case data detection via a specific metric. This metric is developed on surprise adequacy (SA) which has advantages on capture data behaviors. Furthermore, targeting at characteristics of corner case data, three modifications on distanced-based SA are developed for classification applications in this paper. Consequently, through the experiment analysis on MNIST data and industrial data, the feasibility and usefulness of the proposed method on corner case data detection are verified.

Vicent Sanz Marco, Ben Taylor, Zheng Wang et al.

Deep neural networks ( DNNs ) are becoming a key enabling technology for many application domains. However, on-device inference on battery-powered, resource-constrained embedding systems is often infeasible due to prohibitively long inferencing time and resource requirements of many DNNs. Offloading computation into the cloud is often unacceptable due to privacy concerns, high latency, or the lack of connectivity. While compression algorithms often succeed in reducing inferencing times, they come at the cost of reduced accuracy. This paper presents a new, alternative approach to enable efficient execution of DNNs on embedded devices. Our approach dynamically determines which DNN to use for a given input, by considering the desired accuracy and inference time. It employs machine learning to develop a low-cost predictive model to quickly select a pre-trained DNN to use for a given input and the optimization constraint. We achieve this by first off-line training a predictive model, and then using the learned model to select a DNN model to use for new, unseen inputs. We apply our approach to two representative DNN domains: image classification and machine translation. We evaluate our approach on a Jetson TX2 embedded deep learning platform and consider a range of influential DNN models including convolutional and recurrent neural networks. For image classification, we achieve a 1.8x reduction in inference time with a 7.52% improvement in accuracy, over the most-capable single DNN model. For machine translation, we achieve a 1.34x reduction in inference time over the most-capable single model, with little impact on the quality of translation.

Ben Taylor, Vicent Sanz Marco, Willy Wolff et al.

The recent ground-breaking advances in deep learning networks ( DNNs ) make them attractive for embedded systems. However, it can take a long time for DNNs to make an inference on resource-limited embedded devices. Offloading the computation into the cloud is often infeasible due to privacy concerns, high latency, or the lack of connectivity. As such, there is a critical need to find a way to effectively execute the DNN models locally on the devices. This paper presents an adaptive scheme to determine which DNN model to use for a given input, by considering the desired accuracy and inference time. Our approach employs machine learning to develop a predictive model to quickly select a pre-trained DNN to use for a given input and the optimization constraint. We achieve this by first training off-line a predictive model, and then use the learnt model to select a DNN model to use for new, unseen inputs. We apply our approach to the image classification task and evaluate it on a Jetson TX2 embedded deep learning platform using the ImageNet ILSVRC 2012 validation dataset. We consider a range of influential DNN models. Experimental results show that our approach achieves a 7.52% improvement in inference accuracy, and a 1.8x reduction in inference time over the most-capable single DNN model.