Search

The RRAM-based accelerators have become very popular candidates for neural network acceleration due to they perform matrix-vector multiplication in-memory with high storage density and low latency. Many related works have used fixed-precision quantization to achieve the model compression and enhance the tolerance of process variation, but these methods still suffer a large accuracy degradation and poor robustness to nonideal effects. In this work, we propose a crossbar-aware mixed-precision quantization scheme, which enables to search for the optimal precision of each part of the network as a way to improve the accuracy and robustness to noise. First, we introduce a group quantization strategy that can flexibly adjust the group size dynamically according to the crossbar size. Then, we propose a detailed mixed-precision search flow to search for the optimal precision set of the network. Finally, we give a noise injection adaption training method to enhance the tolerance of noise. Experimental results show that our proposed method can improve inference accuracy by at least 2.04% compared to the fixed-precision quantization under the same resource cost. The searched architecture with the highest accuracy most accurate (MA) could achieve an accuracy of 92.39% and a resource saving of 93.30% compared to the full precision model. The searched architecture with the biggest resource savings most efficient (ME) could achieve an accuracy of 91.11% and a resource saving of 95.57% compared to the full precision model. And, the average precision of the ME mixed-precision architecture is only 1.4 bits. Besides, the results show the mixed-precision network with noise adaption training is more robust to noise than the fixed-precision network with noise adaption training. [ABSTRACT FROM AUTHOR]

This letter introduces a novel Fourier-based data augmentation strategy for visible-thermal person re-identification (VT-ReID). Different from some existing methods which are proposed from the perspective of network structure and loss functions, our method aims to fully consider the semantic information from the perspective of data preprocessing. The main hypothesis is that the phase component in the Fourier domain contains high-level semantic information and the amplitude component contains low-level modality awareness information. In order to make the model pay more attention to semantic information learning, we design a simple but effective Fourier-based semantic augmentation (FSA) module, which can be inserted seamlessly into any existing models. Extensive experiments on RegDB and SYSU-MM01 datasets have shown that our proposed method can improve the VT-ReID performance significantly and achieve state-of-the-art performance. [ABSTRACT FROM AUTHOR]

The flexible robot has more advantages over the traditional machine tools in machining complex workpieces like the 3D-printed part which has less residual material for removal. In robot machining, the geometric smoothness of moving joints is vital for enhancing efficiency and accuracy and the tool orientation should be limited in certain regions considering the interfering requirements. As the robot has six DoFs at least, redundant DoFs are observed for robot surface machining, which can be employed to optimize the smooth machining process under the tool orientation requirements. In this paper, the tool posture differential vector along the tool-tip path curve is optimized firstly. The whole trajectory is numerically integrated by the optimized differential vector subsequently. In each step, the minimum 2-norm of the joints’ differential vector along the tool path curve is set as part of the optimization objective to ensure the minimal change of joints. Furthermore, a state-related optimization objective for the differential vector is established to keep the tool orientation away from the boundary of the feasible region. Combining with the two objectives, the tool posture differential vector along the tool-tip path is optimized and the whole machining process is obtained by the numerical integration method. As a case study, the joint trajectories for an inclined butterfly curve are planned by the algorithm and machined by the UR-5 robot with the ROS controller. In addition, a comparative experiment is also provided to verify the effectiveness and optimality of the proposed method. [ABSTRACT FROM AUTHOR]

Many unmixing methods hold the assumption that endmembers correspond to major land covers, but not true for some unmixing tasks where observed minor object signals corresponding to some special types of endmembers are relatively weak. When there exist weak signals that have low intensity potentially caused by subtle mixing abundance fractions regarding the endmembers of minor objects, the traditional unmixing techniques may fail. This article pioneers weak signal scenarios in hyperspectral unmixing using an efficient method called HyperWeak. In particular, HyperWeak involves a sparse nonnegative matrix factorization (NMF) model that contains two main parts, where the unsupervised part estimates the endmember and abundance matrices, and the supervised part ensures the minimal degradation of prior knowledge. To enhance the robustness of the HyperWeak model, this article considers a reweighted sparsity constraint to boost the sparseness of the abundance matrix. For effectively solving optimization problems, a Nesterov’s optimal gradient method (OGM) is used in this article. Experiments conducted on synthetic and real hyperspectral images indicate that HyperWeak can improve the unmixing performances of hyperspectral data in weak signal situations. [ABSTRACT FROM AUTHOR]

Recently, binary neural networks have received increasing interest in object detection community, since compared to the full-precision detection networks, they can reduce the memory and computation requirements significantly due to the efficient XNOR and BITCOUNT operations introduced by binarization. However, the final detection performance of binary detectors always suffers from a severe degradation compared with the full-precision counterparts. In this letter, to achieve a better trade-off between the inference efficiency and the detection performance, we propose to learn a laterally connected binary detector based on the multiple parallel binary subnets with a group sparse regularization term. Firstly, we simply adopt the parallel structure to improve the representation capacity of binary detectors. Secondly, we introduce the dense lateral connections between binary subnets in each convolutional layer to increase the information flows. Finally, to reduce the redundancy of the dense connections, we take the L21 norm as the regularization term to optimize the lateral connections in an end-to-end manner. Experimental results on PASCAL VOC and COCO datasets show that our proposed laterally connected binary detector could outperform the other state-of-the-art binary detectors. [ABSTRACT FROM AUTHOR]

The vector neural network (VNN) based on memristor has tremendous potential in applications such as electronic reconnaissance, medical diagnosis, and speech processing. However, the VNNs that encompass a huge amount of multiply-accumulate (MAC) operations often acquire network weights through massive numerical calculations with high precision, which results in a heavy consumption of energy and computing resource. Nevertheless, the resistance states of existing memristive devices can hardly meet the high-precision regulation requirements that restricts the application of VNN. We propose a binary memristor based vector-type back propagation (BMVTBP) architecture that integrates the advantages of low-precision memristive devices and VNN. The core function of this architecture is to realize the low-precision weights of three states by employing binarized memristive synapses, and further construct the positive and negative synaptic arrays to implement the MAC operations of interval data. Simulations verify the identification performance of the BMVTBP architecture on the emitter library. The ensuing results demonstrate that the identification rates exceed 96% and 87% for the interval-value and scalar-value noisy emitter samples, respectively, with an architecture requirement of only 1920 memristive devices and an energy consumption of about 2.43e-10 J. [ABSTRACT FROM AUTHOR]

We address the problem of action detection in continuous untrimmed video streams, based on the two-stage framework: one stage for action proposals generation and the other for proposals classification and refinement. The context features inside and outside a candidate region (proposal) are critical for classification in action detection. Therefore, effective integration of these features with different scales has become a fundamental problem. We contend that different action instances and candidate proposals may need different context features. To address this issue, we present a novel multiple scales based context-aware net (MSCA-Net) to effectively classify the action proposals for action detection in this paper. For each candidate action proposal, MSCA-Net takes its multiple regions with different temporal scales as input and then generates suitable context features. Based on the “candidate-control” mechanism of LSTM, the proposed MSCA-Net specially adopts the two-branch structure: Branch1 generates multi-scale context features for each candidate proposal, whereas Branch2 utilizes the context-aware gate function to control the message passing. Extensive experiments on THUMOS’14, Charades daily and ActivityNet action detection datasets, demonstrate the effectiveness of the designed structure and show how these context features influence the detection results. [ABSTRACT FROM AUTHOR]

How to mine the gene regulatory relationship and construct gene regulatory network (GRN) is of utmost interest within the whole biological community, however, which has been consistently a challenging problem since the tremendous complexity in cellular systems. In present work, we construct gene regulatory network using an improved three-phase dependency analysis algorithm (TPDA) Bayesian network learning method, which includes the steps of Drafting, Thickening, and Thinning. In order to solve the problem of learning result is not reliable due to the high order conditional independence test, we use the entropy estimation approach of Gaussian kernel probability density estimator to calculate the (conditional) mutual information between genes. The experiment on the public benchmark data sets show the improved method outperforms the other nine kinds of Bayesian network learning methods when to process the data with large sample size, with small number of discrete values, and the frequency of different discrete values is about same. In addition, the improved TPDA method was further applied on a real large gene expression data set on RNA-seq from a global collection with 368 elite maize inbred lines. Experiment results show it performs better than the original TPDA method and the other nine kinds of Bayesian network learning algorithms significantly. [ABSTRACT FROM AUTHOR]