A discrete convolutional network for entity relation extraction

Relation extraction independently verifies all entity pairs in a sentence to identify predefined relationships between named entities. Because these entity pairs share the same contextual features of a sentence, they lead to a complicated semantic structure. To distinguish semantic expressions between relation instances, manually designed rules or elaborate deep architectures are usually applied to learn task-relevant representations. In this paper, a discrete convolutional network is proposed to incorporate discrete linguistic interactions and deep feature weighting. This network applies a discretization strategy to fix parameters of convolutional kernels into ternary values. Then, these discretized kernels are used to learn discrete semantic structures from vectorized token representations. Our approach leverages the ability of discrete CNNs to capture discrete linguistic patterns of a sentence, thereby maintaining model expressiveness and improving performance in the relation extraction task. Furthermore, our method has the advantages of reducing the overfitting problem caused by depending on prior knowledge and decreasing the computational complexity by reducing the number of trainable parameters. Our model is evaluated on five widely used benchmark datasets. It achieves state-of-the-art performance, outperforming all compared related works. Experimental results also demonstrate that, compared with traditional CNN networks, it achieves an average improvement of 14.66% in F1-score and accelerates training by an average of 17.46%, highlighting the efficiency and effectiveness of our model in the relation extraction task.

Binary Neural Networks in FPGAs: Architectures, Tool Flows and Hardware Comparisons

Binary neural networks (BNNs) are variations of artificial/deep neural network (ANN/DNN) architectures that constrain the real values of weights to the binary set of numbers {-1,1}. By using binary values, BNNs can convert matrix multiplications into bitwise operations, which accelerates both training and inference and reduces hardware complexity and model sizes for implementation. Compared to traditional deep learning architectures, BNNs are a good choice for implementation in resource-constrained devices like FPGAs and ASICs. However, BNNs have the disadvantage of reduced performance and accuracy because of the tradeoff due to binarization. Over the years, this has attracted the attention of the research community to overcome the performance gap of BNNs, and several architectures have been proposed. In this paper, we provide a comprehensive review of BNNs for implementation in FPGA hardware. The survey covers different aspects, such as BNN architectures and variants, design and tool flows for FPGAs, and various applications for BNNs. The final part of the paper gives some benchmark works and design tools for implementing BNNs in FPGAs based on established datasets used by the research community.

Feature Extraction of Ancient Chinese Characters Based on Deep Convolution Neural Network and Big Data Analysis

In recent years, deep learning has made good progress and has been applied to face recognition, video monitoring, image processing, and other fields. In this big data background, deep convolution neural network has also received more and more attention. In order to extract the ancient Chinese characters effectively, the paper will discuss the structure model, pool process, and network training of deep convolution neural network and compare the algorithm with the traditional machine learning algorithm. The results show that the accuracy and recall rate of the Chinese characters in the plaque of Ming Dynasty can reach the peak, 81.38% and 81.31%, respectively. When the number of training samples increases to 50, the recognition rate of MFA is 99.72%, which is much higher than other algorithms. This shows that the algorithm based on deep convolution neural network and big data analysis has excellent performance and can effectively identify the Chinese characters under different dynasties, different sample sizes, and different interference factors, which can provide a powerful reference for the extraction of ancient Chinese characters.

An improved object detection algorithm based on multi-scaled and deformable convolutional neural networks

Object detection methods aim to identify all target objects in the target image and determine the categories and position information in order to achieve machine vision understanding. Numerous approaches have been proposed to solve this problem, mainly inspired by methods of computer vision and deep learning. However, existing approaches always perform poorly for the detection of small, dense objects, and even fail to detect objects with random geometric transformations. In this study, we compare and analyse mainstream object detection algorithms and propose a multi-scaled deformable convolutional object detection network to deal with the challenges faced by current methods. Our analysis demonstrates a strong performance on par, or even better, than state of the art methods. We use deep convolutional networks to obtain multi-scaled features, and add deformable convolutional structures to overcome geometric transformations. We then fuse the multi-scaled features by up sampling, in order to implement the final object recognition and region regress. Experiments prove that our suggested framework improves the accuracy of detecting small target objects with geometric deformation, showing significant improvements in the trade-off between accuracy and speed.

Modeling global geometric spatial information for rotation invariant classification of satellite images

The classification of high-resolution satellite images is an open research problem for computer vision research community. In last few decades, the Bag of Visual Word (BoVW) model has been used for the classification of satellite images. In BoVW model, an orderless histogram of visual words without any spatial information is used as image signature. The performance of BoVW model suffers due to this orderless nature and addition of spatial clues are reported beneficial for scene and geographical classification of images. Most of the image representations that can compute image spatial information as are not invariant to rotations. A rotation invariant image representation is considered as one of the main requirement for satellite image classification. This paper presents a novel approach that computes the spatial clues for the histograms of BoVW model that is robust to the image rotations. The spatial clues are calculated by computing the histograms of orthogonal vectors. This is achieved by calculating the magnitude of orthogonal vectors between Pairs of Identical Visual Words (PIVW) relative to the geometric center of an image. The comparative analysis is performed with recently proposed research to obtain the best spatial feature representation for the satellite imagery. We evaluated the proposed research for image classification using three standard image benchmarks of remote sensing. The results and comparisons conducted to evaluate this research show that the proposed approach performs better in terms of classification accuracy for a variety of datasets based on satellite images.

Visual complexity modelling based on image features fusion of multiple kernels

Humans' perception of visual complexity is often regarded as one of the key principles of aesthetic order, and is intimately related to the physiological, neurological and, possibly, psychological characteristics of the human mind. For these reasons, creating accurate computational models of visual complexity is a demanding task. Building upon on previous work in the field (Forsythe et al., 2011; Machado et al., 2015) we explore the use of Machine Learning techniques to create computational models of visual complexity. For that purpose, we use a dataset composed of 800 visual stimuli divided into five categories, describing each stimulus by 329 features based on edge detection, compression error and Zipf's law. In an initial stage, a comparative analysis of representative state-of-the-art Machine Learning approaches is performed. Subsequently, we conduct an exhaustive outlier analysis. We analyze the impact of removing the extreme outliers, concluding that Feature Selection Multiple Kernel Learning obtains the best results, yielding an average correlation to humans' perception of complexity of 0.71 with only twenty-two features. These results outperform the current state-of-the-art, showing the potential of this technique for regression.