Search

Program execution on intermittently powered Internet-of-Things (IoT) devices must ensure forward progress in the presence of frequent power failures. A general solution is intermittent computing, by which the program states are frequently checkpointed to non-volatile memory (NVM) so that once a power failure occurs, the program can restart from the latest checkpoint after the system energy regains. However, executing a deep neural network (DNN) inference program in an intermittent way has a big problem. During the execution, an inference program will generate large-volume feature maps (as part of the program states), and checkpointing the feature maps to NVM will incur significant time and energy overhead and thus reduce the inference efficiency. This paper proposes an approach to reduce the amount of feature map writing in intermittent DNN inference. The main idea is to partition the inference task into several slices and execute each slice in a depth-first manner so that the intermediate feature maps during the inference of each slice do not need to be written to NVM. Extensive experiments have been conducted, which show that the proposed approach can significantly reduce the amount of NVM writing, and a maximum of 1.965 speedup of the total inference time is achieved compared to the state-of-the-art approach.

Internet-of-Things (IoT) devices are becoming both intelligent and green. On the one hand, Deep Neural Network (DNN) compression techniques make it possible to run deep learning applications on devices equipped with low-end microcontrollers (MCUs). By performing deep learning on IoT devices, in-situ decision-making can be made, which can improve the responsiveness of such devices to the environment and reduce data uploading to edge servers or clouds to save valuable network bandwidth. On the other hand, many IoT devices in the future will be powered by energy harvesters instead of batteries to reduce environmental pollution and achieve permanent service free of battery maintenance. As the energy output of energy harvesters is tiny and unstable, energy harvesting IoT (EH-IoT) devices will experience frequent power failures during their execution, making the software task hard to progress. The deep learning tasks running on such devices must face this challenge and, at the same time, ensure satisfactory execution efficiency. We believe deploying deep learning on EH-IoT devices that execute intermittently will be a challenging yet promising research direction. To motivate research in this direction, this paper summarizes existing solutions and provides an in-depth discussion of future challenges that deserve further investigation. With IoT devices becoming more intelligent and green, DNN inference on EH-IoT devices will generate a much more significant impact in the future in academia and industry.

Background: Today's machine-learning based dermatologic research has largely focused on pigmented/non-pigmented lesions concerning skin cancers. However, studies on machine-learning-aided diagnosis of depigmented non-melanocytic lesions, which are more difficult to diagnose by unaided eye, are very few.Objective: We aim to assess the performance of deep learning methods for diagnosing vitiligo by deploying Convolutional Neural Networks (CNNs) and comparing their diagnosis accuracy with that of human raters with different levels of experience.Methods: A Chinese in-house dataset (2,876 images) and a world-wide public dataset (1,341 images) containing vitiligo and other depigmented/hypopigmented lesions were constructed. Three CNN models were trained on close-up images in both datasets. The results by the CNNs were compared with those by 14 human raters from four groups: expert raters (>10 years of experience), intermediate raters (5–10 years), dermatology residents, and general practitioners. F1 score, the area under the receiver operating characteristic curve (AUC), specificity, and sensitivity metrics were used to compare the performance of the CNNs with that of the raters.Results: For the in-house dataset, CNNs achieved a comparable F1 score (mean [standard deviation]) with expert raters (0.8864 [0.005] vs. 0.8933 [0.044]) and outperformed intermediate raters (0.7603 [0.029]), dermatology residents (0.6161 [0.068]) and general practitioners (0.4964 [0.139]). For the public dataset, CNNs achieved a higher F1 score (0.9684 [0.005]) compared to the diagnosis of expert raters (0.9221 [0.031]).Conclusion: Properly designed and trained CNNs are able to diagnose vitiligo without the aid of Wood's lamp images and outperform human raters in an experimental setting.

With the wide-spread adoption of electric vehicles (EVs), introducing solar energy in building EV charging stations is promising as it can reduce carbon emissions and improve air quality. The main challenges are how to decide where to build solar assisted charging stations in a city and how to size the charging stations, as the decision is affected by a broad range of factors, such as construction cost, solar energy fluctuation, and user requirements. This paper proposed an approach to efficiently decide the locations and sizes of solar energy assisted charging stations for an urban area. Experiments are conducted on real EV history data from 297 users of an EV leasing company. The results show that the proposed method can produce high quality decisions within reasonable computation time. The work of this paper will provide important information for decision makers to integrate solar energy into the EV charging infrastructure.

The last level cache (LLC) has significant impact to system performance on modern multi-core processors. But as cache sizes reach several megabytes and more, the overhead of exploring performance on LLC greatly increases as well. To improve the efficiency of performance analysis, we propose a set-sampling-based cache profiling model for the performance analysis on multi-core LLC. We first explore the memory access distributions on LLC by developing a low-overhead stress-application-based method. The results show that memory access distributions can be approximated by Gaussian distribution function. Based on this observation, a Gaussian-distribution-based set sampling model is proposed which can predict program performance with limited representative samples. We evaluate our model on a contemporary multi-core machine and show that 1) the proposed method can precisely predict program performance on LLC under different contention intensities and 2) our method can achieve similar precision with less samples compared to widely adopted set sampling methods such as the random sampling and the continuous address sampling.