Recurrent Neural Networks (RNNs) and their variants, such as Long-Short Term Memory (LSTM) networks, and Gated Recurrent Unit (GRU) networks, have achieved promising performance in sequential data modeling. The hidden layers in RNNs can be regarded as the memory units, which are helpful in storing information in sequential contexts. However, when dealing with high dimensional input data, such as video and text, the input-to-hidden linear transformation in RNNs brings high memory usage and huge computational cost. This makes the training of RNNs very difficult. To address this challenge, we propose a novel compact LSTM model, named as TR-LSTM, by utilizing the low-rank tensor ring decomposition (TRD) to reformulate the input-to-hidden transformation. Compared with other tensor decomposition methods, TR-LSTM is more stable. In addition, TR-LSTM can complete an end-to-end training and also provide a fundamental building block for RNNs in handling large input data. Experiments on real-world action recognition datasets have demonstrated the promising performance of the proposed TR-LSTM compared with the tensortrain LSTM and other state-of-the-art competitors.

Discrimination power analysis (DPA) is a statistical analysis combining discrimination concept with discrete cosine transform coefficients (DCTCs) properties. Unfortunately there is not a uniform and effective criterion to optimize the shape and size of premasking window on which the effect of DPA excessively relies. Proper premasking is an auxiliary process to select the feature coefficients that have more discrimination power (DP). Dynamic weighted DPA (DWDPA) is proposed in this paper to enhance the DP of the selected DCTCs without premasking window, in other words, it does not need to optimize the shape and size of premasking window. The DCTCs are adaptively selected according to their discrimination power values (DPVs). More DCTCs with higher DP are preserved. The selected coefficients are normalized and dynamic weighted according to their DPVs. Normalization assures that the DCTCs with large absolute value don't destroy the DP of the other DCTCs that have less absolute value but high DPVs. Dynamic weighting gives larger weights to the DCTCs with larger DPVs which optimizes and enhances the recognition performance. The experimental results on ORL, Yale and PolyU databases show that DWDPA outperforms DPA obviously.

No abstract

Bolted spherical joints are widely used to form space steel structures. The stiffness and load capacity of the structures are affected by the looseness of bolted spherical joint connections in the structures. The looseness of the connections, which can be caused by fabrication error, low modeling accuracy, and “false twist” in the installation process, may negatively impact the load capacity of the structure and even lead to severe accidents. Furthermore, it is difficult to detect bolted spherical joint connection looseness from the outside since the bolts connect spheres with rods together from the inside. Active sensing methods are proposed in this paper to monitor the tightness status of the bolted spherical connection using piezoceramic transducers. A triangle-on-triangle offset grid composed of bolted spherical joints and steel tube bars was fabricated as the specimen and was used to validate the active sensing methods. Lead Zirconate Titanate (PZT) patches were used as sensors and actuators to monitor the bolted spherical joint tightness status. One PZT patch mounted on the central bolted sphere at the upper chord was used as an actuator to generate a stress wave. Another PZT patch mounted on the bar was used as a sensor to detect the propagated waves through the bolted spherical connection. The looseness of the connection can impact the energy of the stress wave propagated through the connection. The wavelet packet analysis and time reversal (TR) method were used to quantify the energy of the transmitted signal between the PZT patches by which the tightness status of the connection can be detected. In order to verify the effectiveness, repeatability, and consistency of the proposed methods, the experiments were repeated six times in different bolted spherical connection positions. The experimental results showed that the wavelet packet analysis and TR method are effective in detecting the tightness status of the connections. The proposed active monitoring method using PZT transducers can monitor the tightness levels of bolted spherical joint connections efficiently and shows its potential to guarantee the safety of space steel structures in construction and service.

No abstract

In the Internet of Things (IoT) equipment used for aquaculture is often deployed in outdoor ponds located in remote areas. Faults occur frequently in these tough environments and the staff generally lack professional knowledge and pay a low degree of attention in these areas. Once faults happen, expert personnel must carry out maintenance outdoors. Therefore, this study presents an intelligent method for fault diagnosis based on fault tree analysis and a fuzzy neural network. In the proposed method, first, the fault tree presents a logic structure of fault symptoms and faults. Second, rules extracted from the fault trees avoid duplicate and redundancy. Third, the fuzzy neural network is applied to train the relationship mapping between fault symptoms and faults. In the aquaculture IoT, one fault can cause various fault symptoms, and one symptom can be caused by a variety of faults. Four fault relationships are obtained. Results show that one symptom-to-one fault, two symptoms-to-two faults, and two symptoms-to-one fault relationships can be rapidly diagnosed with high precision, while one symptom-to-two faults patterns perform not so well, but are still worth researching. This model implements diagnosis for most kinds of faults in the aquaculture IoT.

Shearers play an important role in fully mechanized coal mining face and accurately identifying their cutting pattern is very helpful for improving the automation level of shearers and ensuring the safety of coal mining. The least squares support vector machine (LSSVM) has been proven to offer strong potential in prediction and classification issues, particularly by employing an appropriate meta-heuristic algorithm to determine the values of its two parameters. However, these meta-heuristic algorithms have the drawbacks of being hard to understand and reaching the global optimal solution slowly. In this paper, an improved fly optimization algorithm (IFOA) to optimize the parameters of LSSVM was presented and the LSSVM coupled with IFOA (IFOA-LSSVM) was used to identify the shearer cutting pattern. The vibration acceleration signals of five cutting patterns were collected and the special state features were extracted based on the ensemble empirical mode decomposition (EEMD) and the kernel function. Some examples on the IFOA-LSSVM model were further presented and the results were compared with LSSVM, PSO-LSSVM, GA-LSSVM and FOA-LSSVM models in detail. The comparison results indicate that the proposed approach was feasible, efficient and outperformed the others. Finally, an industrial application example at the coal mining face was demonstrated to specify the effect of the proposed system.

Backdoor attacks represent a serious threat to neural network models. A backdoored model will misclassify the trigger-embedded inputs into an attacker-chosen target label while performing normally on other benign inputs. There are already numerous works on backdoor attacks on neural networks, but only a few works consider graph neural networks (GNNs). As such, there is no intensive research on explaining the impact of trigger injecting position on the performance of backdoor attacks on GNNs.To bridge this gap, we conduct an experimental investigation on the performance of backdoor attacks on GNNs. We apply two powerful GNN explainability approaches to select the optimal trigger injecting position to achieve two attacker objectives -high attack success rate and low clean accuracy drop. Our empirical results on benchmark datasets and state-of-the-art neural network models demonstrate the proposed method's effectiveness in selecting trigger injecting position for backdoor attacks on GNNs. For instance, on the node classification task, the backdoor attack with trigger injecting position selected by GraphLIME reaches over 84% attack success rate with less than 2.5% accuracy drop. CCS CONCEPTS• Security and privacy → Software and application security;