查看更多>>摘要：? 2021 Elsevier LtdThe prediction of crowd flows is an important urban computing issue whose purpose is to predict the future number of incoming and outgoing people in regions. Measuring the complicated spatial–temporal dependencies with external factors, such as weather conditions and surrounding point-of-interest (POI) distribution is the most difficult aspect of predicting crowd flows movement. To overcome the above issue, this paper advises a unified dynamic deep spatio-temporal neural network model based on convolutional neural networks and long short-term memory, termed as (DHSTNet) to simultaneously predict crowd flows in every region of a city. The DHSTNet model is made up of four separate components: a recent, daily, weekly, and an external branch component. Our proposed approach simultaneously assigns various weights to different branches and integrates the four properties’ outputs to generate final predictions. Moreover, to verify the generalization and scalability of the proposed model, we apply a Graph Convolutional Network (GCN) based on Long Short Term Memory (LSTM) with the previously published model, termed as GCN-DHSTNet; to capture the spatial patterns and short-term temporal features; and to illustrate its exceptional accomplishment in predicting the traffic crowd flows. The GCN-DHSTNet model not only depicts the spatio-temporal dependencies but also reveals the influence of different time granularity, which are recent, daily, weekly periodicity and external properties, respectively. Finally, a fully connected neural network is utilized to fuse the spatio-temporal features and external properties together. Using two different real-world traffic datasets, our evaluation suggests that the proposed GCN-DHSTNet method is approximately 7.9%–27.2% and 11.2%–11.9% better than the AAtt-DHSTNet method in terms of RMSE and MAPE metrics, respectively. Furthermore, AAtt-DHSTNet outperforms other state-of-the-art methods.
查看更多>>摘要：? 2021 Elsevier LtdIntegrating multi-scale predictions has become a mainstream paradigm in edge detection. However, most existing methods mainly focus on effective feature extraction and multi-scale feature fusion while ignoring the low learning capacity in fine-level branches, limiting the overall fusion performance. In light of this, we propose a novel Fine-scale Corrective Learning Net (FCL-Net) that exploits semantic information from deep layers to facilitate fine-scale feature learning. FCL-Net mainly consists of a Top-down Attentional Guiding (TAG) and a Pixel-level Weighting (PW) module. TAG module adopts semantic attentional cues from coarse-scale prediction into guiding the fine-scale branches by learning a top-down LSTM. PW module treats the contribution of each spatial location independently and promote fine-level branches to detect detailed edges with high confidence. Experiments on three benchmark datasets, i.e., BSDS500, Multicue, and BIPED, show that our approach significantly outperforms the baseline and achieves a competitive ODS F-measure of 0.826 on the BSDS500 benchmark. The source code and models are publicly available at https://github.com/DREAMXFAR/FCL-Net.
查看更多>>摘要：? 2021 Elsevier LtdLearning complex tasks from scratch is challenging and often impossible for humans as well as for artificial agents. Instead, a curriculum can be used, which decomposes a complex task – the target task – into a sequence of source tasks. Each source task is a simplified version of the next source task with increasing complexity. Learning then occurs gradually by training on each source task while using knowledge from the curriculum's prior source tasks. In this study, we present a new algorithm that combines curriculum learning with Hindsight Experience Replay (HER), to learn sequential object manipulation tasks for multiple goals and sparse feedback. The algorithm exploits the recurrent structure inherent in many object manipulation tasks and implements the entire learning process in the original simulation without adjusting it to each source task. We test our algorithm on three challenging throwing tasks in simulation and show significant improvements compared to vanilla-HER.
查看更多>>摘要：? 2021 Elsevier LtdReinforcement learning methods have recently been very successful at performing complex sequential tasks like playing Atari games, Go and Poker. These algorithms have outperformed humans in several tasks by learning from scratch, using only scalar rewards obtained through interaction with their environment. While there certainly has been considerable independent innovation to produce such results, many core ideas in reinforcement learning are inspired by phenomena in animal learning, psychology and neuroscience. In this paper, we comprehensively review a large number of findings in both neuroscience and psychology that evidence reinforcement learning as a promising candidate for modeling learning and decision making in the brain. In doing so, we construct a mapping between various classes of modern RL algorithms and specific findings in both neurophysiological and behavioral literature. We then discuss the implications of this observed relationship between RL, neuroscience and psychology and its role in advancing research in both AI and brain science.
查看更多>>摘要：? 2021 Elsevier LtdThe networked systems are booming in multi-disciplines, including the industrial engineering system, the social system, and so on. The network structure is a prerequisite for the understanding and exploration of networked systems. However, the network structure is always unknown in practice, thus, it is significant yet challenging to investigate the inference of network structure. Although some model-based methods and data-driven methods, such as the phase-space based method and the compressive sensing based method, have investigated the structure inference tasks, they were time-consuming due to the greedy iterative optimization procedure, which makes them difficult to satisfy real-time structure inference requirements. Although the reconstruction time of L1 and other methods is short, the reconstruction accuracy is very low. Inspired by the powerful representation ability and time efficiency for the structure inference with the deep learning framework, a novel synergy method combines the deep residual network and fully connected layer network to solve the network structure inference task efficiently and accurately. This method perfectly solves the problems of long reconstruction time and low accuracy of traditional methods. Moreover, the proposed method can also fulfill the inference task of large scale complex network, which further indicates the scalability of the proposed method.
查看更多>>摘要：? 2021 Elsevier LtdClipping on learning rates in Adam leads to an effective stochastic algorithm—AdaBound. In spite of its effectiveness in practice, convergence analysis of AdaBound has not been fully explored, especially for non-convex optimization. To this end, we address the convergence of the last individual output of AdaBound for non-convex stochastic optimization problems, which is called individual convergence. We prove that, with the iteration of the AdaBound, the cost function converges to a finite value and the corresponding gradient converges to zero. The novelty of this proof is that the convergence conditions on the bound functions and momentum factors are much more relaxed than the existing results, especially when we remove the monotonicity and convergence of the bound functions, and only keep their boundedness. The momentum factors can be fixed to be constant, without the restriction of monotonically decreasing. This provides a new perspective on understanding the bound functions and momentum factors of AdaBound. At last, numerical experiments are provided to corroborate our theory and show that the convergence of AdaBound extends to more general bound functions.
查看更多>>摘要：? 2021 Elsevier LtdEmotion classification based on neurophysiology signals has been a challenging issue in the literature. Recent neuroscience findings suggest that brain network structure underlying the different emotions provides a window in understanding human affection. In this paper, we propose a novel method to capture the distinct minimum spanning tree (MST) topology underpinning the different emotions. Specifically, we propose a hierarchical aggregation-based graph neural network to investigate the MST structure in emotion recognition. Extensive experiments on the public available DEAP dataset demonstrate the superior performance of the model in emotion classification as compared to existing methods. In addition, the results show that the theta, lower beta and gamma frequency band network information are more sensitive to emotions, suggesting a multi-frequency interaction in emotion processing.
查看更多>>摘要：? 2021 Elsevier LtdIn this article, we discuss bipartite fixed-time synchronization for fractional-order signed neural networks with discontinuous activation patterns. The Filippov multi-map is used to convert the fixed-time stability of the fractional-order general solution into the zero solution of the fractional-order differential inclusions. On the Caputo fractional-order derivative, Lyapunov-Krasovskii functional is proved to possess the indefinite fractional derivatives for fixed-time stability of fragmentary discontinuous systems. Furthermore, the fixed-time stability of the fractional-order discontinuous system is achieved as well as an estimate of the new settling time. The discontinuous controller is designed for the delayed fractional-order discontinuous signed neural networks with antagonistic interactions and new conditions for permanent fixed-time synchronization of these networks with antagonistic interactions are also provided, as well as the settling time for permanent fixed-time synchronization. Two numerical simulation results are presented to demonstrate the effectiveness of the main results
查看更多>>摘要：? 2021 Elsevier LtdDriven by industrial big data and intelligent manufacturing, deep learning approaches have flourished and yielded impressive achievements in the community of machine fault diagnosis. Nevertheless, current diagnosis models trained on a specific dataset seldom work well on other datasets due to the domain discrepancy. Recently, adversarial domain adaptation-based approaches have become an emerging and compelling tool to address this issue. Nonetheless, existing methods still have a shortcoming since they cannot guarantee sufficient feature similarity between the source domain and the target domain after adaptation, resulting in unguaranteed performance. To this end, a Cycle-consistent Adversarial Adaptation Network (CAAN) is advanced to realize more effective fault diagnosis of machinery. In CAAN, specifically, an adversarial game formed by the feature extractor and the domain discriminator is constructed to induce transferable feature learning. Meanwhile, the feature translators and discriminators between source and target domains are further designed to build a more comprehensive cycle-consistent generative adversarial constrain, which can more reliably ensure domain-invariant and class-separate characteristics of learned features. Extensive experiments constructed on three datasets from different mechanical systems demonstrate the effectiveness and superiority of CAAN.
查看更多>>摘要：? 2021 Elsevier LtdFinancial market predictions represent a complex problem. Most prediction systems work with the term time window, which is represented by exchange rate values of a real financial commodity. Such values (time window) provide the base for prediction of future values. Real situations, however, prove that prediction of only a single time-series trend is insufficient. This article aims at suggesting a novelty and unconventional approach based on the use of several neural networks predicting probable courses of a future trend defined in a prediction time window. The basis of the proposed approach is a suitable representation of the training-set input data into the neural networks. It uses selected FFT coefficients as well as robust output indicators based on a histogram of the predicted course of the selected currency pair. At the same time, the given currency pair enters the prediction in a combination with another three mutually interconnected currency pairs. A significant output of the articles is, apart from the proposed methodology, confirmation that the Elliott wave theory is beneficial in the trading environment and provides a substantial profit compared with conventional prediction techniques. That was proved in the performed experimental study.
NSTL
Elsevier