High-performance deep neural network (DNN)-based systems are in high demand in edge environments. Due to its high computational complexity, it is challenging to deploy DNNs on edge devices with strict limitations on computational resources. In this paper, we derive a compact while highly-accurate DNN model, termed dsODENet, by combining recently-proposed parameter reduction techniques: Neural ODE (Ordinary Differential Equation) and DSC (Depthwise Separable Convolution). Neural ODE exploits a similarity between ResNet and ODE, and shares most of weight parameters among multiple layers, which greatly reduces the memory consumption. We apply dsODENet to a domain adaptation as a practical use case with image classification datasets. We also propose a resource-efficient FPGA-based design for dsODENet, where all the parameters and feature maps except for pre- and post-processing layers can be mapped onto on-chip memories. It is implemented on Xilinx ZCU104 board and evaluated in terms of domain adaptation accuracy, inference speed, FPGA resource utilization, and speedup rate compared to a software counterpart. The results demonstrate that dsODENet achieves comparable or slightly better domain adaptation accuracy compared to our baseline Neural ODE implementation, while the total parameter size without pre- and post-processing layers is reduced by 54.2% to 79.8%. Our FPGA implementation accelerates the inference speed by 23.8 times.
We developed a PYNQ cluster that consists of economical Zynq boards, called M-KUBOS, that are interconnected through low-cost high-performance GTH serial links. For the software environment, we employed the PYNQ open-source software platform. The PYNQ cluster is anticipated to be a multi-access edge computing (MEC) server for 5G mobile networks. We implemented the ResNet-50 inference accelerator on the PYNQ cluster for image recognition of MEC applications. By estimating the execution time of each ResNet-50 layer, layers of ResNet-50 were divided into multiple boards so that the execution time of each board would be as equal as possible for efficient pipeline processing. Owing to the PYNQ cluster in which FPGAs were directly connected by high-speed serial links, stream processing without network bottlenecks and pipeline processing between boards were readily realized. The implementation on 4 boards achieved 292 GOPS performance, 75.1 FPS throughput, and 7.81 GOPS/W power efficiency. It achieved 17 times faster speed and 130 times more power efficiency compared to the implementation on the CPU, and 5.8 times more power efficiency compared to the implementation on the GPU.
The issue of a low minority class identification rate caused by data imbalance in anomaly detection tasks is addressed by the proposal of a GAN-SR-based intrusion detection model for industrial control systems. First, to correct the imbalance of minority classes in the dataset, a generative adversarial network (GAN) processes the dataset to reconstruct new minority class training samples accordingly. Second, high-dimensional feature extraction is completed using stacked asymmetric depth self-encoder to address the issues of low reconstruction error and lengthy training times. After that, a random forest (RF) decision tree is built, and intrusion detection is carried out using the features that SNDAE retrieved. According to experimental validation on the UNSW-NB15, SWaT and Gas Pipeline datasets, the GAN-SR model outperforms SNDAE-SVM and SNDAE-KNN in terms of detection performance and stability.
The total number of solar power-producing facilities whose Feed-in Tariff (FIT) Program-based ten-year contracts will expire by 2023 is expected to reach approximately 1.65 million in Japan. If the facilities that produce or consume renewable energy would increase to reach a large number, e.g., two million, blockchain would not be capable of processing all the transactions. In this work, we propose a blockchain-based electricity-tracking platform for renewable energy, called ‘ZGridBC,’ which consists of mutually cooperative two novel decentralized schemes to solve scalability, storage cost, and privacy issues at the same time. One is the electricity production resource management, which is an efficient data management scheme that manages electricity production resources (EPRs) on the blockchain by using UTXO tokens extended to two-dimension (period and electricity amount) to prevent double-spending. The other is the electricity-tracking proof, which is a massive data aggregation scheme that significantly reduces the amount of data managed on the blockchain by using zero-knowledge proof (ZKP). Thereafter, we illustrate the architecture of ZGridBC, consider its scalability, security, and privacy, and illustrate the implementation of ZGridBC. Finally, we evaluate the scalability of ZGridBC, which handles two million electricity facilities with far less cost per environmental value compared with the price of the environmental value proposed by METI (=0.3 yen/kWh).
In industry, automatic speech recognition has come to be a competitive feature for embedded products with poor hardware resources. In this work, we propose a tiny end-to-end speech recognition model that is lightweight and easily deployable on edge platforms. First, instead of sophisticated network structures, such as recurrent neural networks, transformers, etc., the model we propose mainly uses convolutional neural networks as its backbone. This ensures that our model is supported by most software development kits for embedded devices. Second, we adopt the basic unit of MobileNet-v3, which performs well in computer vision tasks, and integrate the features of the hidden layer at different scales, thus compressing the number of parameters of the model to less than 1 M and achieving an accuracy greater than that of some traditional models. Third, in order to further reduce the CPU computation, we directly extract acoustic representations from 1-dimensional speech waveforms and use a self-supervised learning approach to encourage the convergence of the model. Finally, to solve some problems where hardware resources are relatively weak, we use a prefix beam search decoder to dynamically extend the search path with an optimized pruning strategy and an additional initialism language model to capture the probability of between-words in advance and thus avoid premature pruning of correct words. In our experiments, according to a number of evaluation categories, our end-to-end model outperformed several tiny speech recognition models used for embedded devices in related work.
In CRYPTO 2019, Gohr first introduced the deep learning method to cryptanalysis for SPECK32/64. A differential-neural distinguisher was obtained using ResNet neural network. Zhang et al. used multiple parallel convolutional layers with different kernel sizes to capture information from multiple dimensions, thus improving the accuracy or obtaining a more round of distinguisher for SPECK32/64 and SIMON32/64. Inspired by Zhang's work, we apply the network structure to other ciphers. We not only improve the accuracy of the distinguisher, but also increase the number of rounds of the distinguisher, that is, distinguish more rounds of ciphertext and random number for DES, Chaskey and PRESENT.
This paper proposes an enhanced model of Random Projection Outlyingness (RPO) for unsupervised outlier detection. When datasets have multiple modalities, the RPOs have frequent detection errors. The proposed model deals with this problem via unsupervised clustering and a local score weighting. The experimental results demonstrate that the proposed model outperforms RPO and is comparable with other existing unsupervised models on benchmark datasets, in terms of in terms of Area Under the Curves (AUCs) of Receiver Operating Characteristic (ROC).
Since the dark channel prior (DCP)-based dehazing method is ineffective in the sky area and will cause the problem of too dark and color distortion of the image, we propose a novel dehazing method based on sky area segmentation and image fusion. We first segment the image according to the characteristics of the sky area and non-sky area of the image, then estimate the atmospheric light and transmission map according to the DCP and correct them, and then fuse the original image after the contrast adaptive histogram equalization to improve the details information of the image. Experiments illustrate that our method performs well in dehazing and can reduce image distortion.
Single image deraining is an ill-posed problem which also has been a long-standing issue. In past few years, convolutional neural network (CNN) methods almost dominated the computer vision and achieved considerable success in image deraining. Recently the Swin Transformer-based model also showed impressive performance, even surpassed the CNN-based methods and became the state-of-the-art on high-level vision tasks. Therefore, we attempt to introduce Swin Transformer to deraining tasks. In this paper, we propose a deraining model with two sub-networks. The first sub-network includes two branches. Rain Recognition Network is a Unet with the Swin Transformer layer, which works as preliminarily restoring the background especially for the location where rain streaks appear. Detail Complement Network can extract the background detail beneath the rain streak. The second sub-network which called Refine-Unet utilizes the output of the previous one to further restore the image. Through experiments, our network achieves improvements on single image deraining compared with the previous Transformer research.
Convolutional Neural Networks (CNNs) have shown remarkable performance in image recognition tasks. In this letter, we propose a new CNN model called the EnsNet which is composed of one base CNN and multiple Fully Connected SubNetworks (FCSNs). In this model, the set of feature maps generated by the last convolutional layer in the base CNN is divided along channels into disjoint subsets, and these subsets are assigned to the FCSNs. Each of the FCSNs is trained independent of others so that it can predict the class label of each feature map in the subset assigned to it. The output of the overall model is determined by majority vote of the base CNN and the FCSNs. Experimental results using the MNIST, Fashion-MNIST and CIFAR-10 datasets show that the proposed approach further improves the performance of CNNs. In particular, an EnsNet achieves a state-of-the-art error rate of 0.16% on MNIST.