通道注意网络和模糊划分熵图割的图像去雾

针对雾图成像时变化的场景光及去雾过程中不同雾相关信息在处理上的差异性,提出了通道注意网络和模糊划分熵图割的单幅图像去雾算法。以考虑变化场景光的大气散射物理成像模型为基础,首先使用通道注意的编码解码网络来估计透射率,并在编码器最后及解码器起始处添加通道注意模块,以便为编码器提取的不同雾相关特征图分配不同的权重,准确地计算透射率;然后利用所提出的模糊划分熵图割算法将透射率划分为不同场景光覆盖下的近景、中景、远景,此分割策略将考虑空间相关性的图割算法与模糊划分熵的阈值分割算法相结合,解决了单一阈值分割算法产生的区域误分问题;最后估计场景光和大气光,得到去雾图像。实验结果表明,算法在合成雾图及真实雾图上均有较好的去雾效果。与已有的去雾算法相比,本文算法在峰值信噪比及结构相似性上均有提升,单张图像的平均处理时间为3.9 s。     

注意力视角下媒体报道对股价的影响模型研究

鉴于中国A股市场个人投资者比例相对较高但其理性程度相对较低的实际情况,在拓展Tetlock模型的基础上构建了一个两资产三阶段的理性期望模型。研究发现:首先,被报道股票的价格受到注意力效应的影响,存在正向注意力溢价;其次,注意力交易者对被报道股票的交易量在媒体报道后放大,且存在买卖不平衡性;最后,作为对有限理性的补偿,注意力受媒体报道的影响程度越高,其收益越低。论文在统一的分析框架之内考察了媒体报道对缓解信息不对称和引发投资者注意力效应两方面的影响,有助于在理论上为进一步探索媒体报道对股票市场的影响机理、在实践中为监管层更好地利用媒体力量促进股市健康发展提供理论启示。     

Prediction of Time Series Gene Expression and Structural Analysis of Gene Regulatory Networks Using Recurrent Neural Networks

Methods for time series prediction and classification of gene regulatory networks (GRNs) from gene expression data have been treated separately so far. The recent emergence of attention-based recurrent neural network (RNN) models boosted the interpretability of RNN parameters, making them appealing for the understanding of gene interactions. In this work, we generated synthetic time series gene expression data from a range of archetypal GRNs and we relied on a dual attention RNN to predict the gene temporal dynamics. We show that the prediction is extremely accurate for GRNs with different architectures. Next, we focused on the attention mechanism of the RNN and, using tools from graph theory, we found that its graph properties allow one to hierarchically distinguish different architectures of the GRN. We show that the GRN responded differently to the addition of noise in the prediction by the RNN and we related the noise response to the analysis of the attention mechanism. In conclusion, this work provides a way to understand and exploit the attention mechanism of RNNs and it paves the way to RNN-based methods for time series prediction and inference of GRNs from gene expression data.     

关于注意和记忆的神经动力学机制

利用随机的相变动力学理论研究了一个具有不同相位的神经振子群模型,并考察神经 振子群对刺激信息的处理及神经编码的动态演化. 通过对动力学模型的数值分析,在二维相 空间上描述了神经元集群内不同振子簇发放动作电位时,数密度随时间演化的图像. 数值分 析的结果表明该模型能够用来描述注意和记忆的神经动力学机制,并且证明了只有高维的神 经动力学模型才能更深刻地描述神经元集群的动力学特性,而以往的编码模型丢失了大量有 用的神经信息.     

Young children's recognition of quantitative relations in mathematically unspecified settings

Children have been found to be able to reason about quantitative relations, such as non-symbolic proportions, already by the age of 5 years. However, these studies utilize settings in which children were explicitly guided to notice the mathematical nature of the tasks. This study investigates children's spontaneous recognition of quantitative relations on mathematically unspecified settings. Participants were 86 Finnish-speaking children, ages 5–8. Two video-recorded tasks, in which participants were not guided to notice the mathematical aspects, were used. The tasks could be completed in a number of ways, including by matching quantitative relations, numerosity, or other aspects. Participants’ matching strategies were analyzed with regard to the most mathematically advanced level utilized. There were substantial differences in participants’ use of quantitative relations, numerosity and other aspects in their matching strategies. The results of this novel experimental setting show that investigating children's spontaneous recognition of quantitative relations provides novel insight into children's mathematical thinking and furthers the understanding of how children recognize and utilize mathematical aspects when not explicitly guided to do so.     

Spatial Spread of Visual Attention while Tracking a Moving Object

We conducted three experiments to investigate the spatial spread of visual attention. In Experiment 1, we measured the contrast sensitivities at various locations (spatial sensitivity function) relative to the moving target that the observer attended to track in an attentive tracking display. A probe was presented at a distance from the target at a location randomly chosen from within a certain range. The range of probe presentation location varied to examine whether the observer changes the area of attention to cope with this range. The results show that the probe range influenced the shape of spatial sensitivity function. The change in shape of this function suggests that the observer covers a wider area with attention for large probe ranges than small probe ranges. In the following experiments, we investigated the effect of the distance between the tracking target and a probe at a fixed location relative to the target (Experiment 2), or between the target and the center of a probe range of fixed size (Experiment 3). Since the relative probe location in a session was fixed in the experiments, the observer would pay attention to the target and probe locations independently of the relative distance if he/she could focus attention at multiple locations. Spatial sensitivity functions obtained in Experiments 2 and 3 showed that this was not the case. In both experiments the sensitivity to the probe decreased with increase in the relative distance as in Experiment 1, where the probe was presented at a location randomly chosen within each range. This indicates that attention cannot be divided among multiple locations, at least under the present experimental conditions. We will discuss a possible interpretation of the present results with a limited attentional resource and its spatial distribution.     

Spatiotemporal Transformer Neural Network for Time-Series Forecasting

Predicting high-dimensional short-term time-series is a difficult task due to the lack of sufficient information and the curse of dimensionality. To overcome these problems, this study proposes a novel spatiotemporal transformer neural network (STNN) for efficient prediction of short-term time-series with three major features. Firstly, the STNN can accurately and robustly predict a high-dimensional short-term time-series in a multi-step-ahead manner by exploiting high-dimensional/spatial information based on the spatiotemporal information (STI) transformation equation. Secondly, the continuous attention mechanism makes the prediction results more accurate than those of previous studies. Thirdly, we developed continuous spatial self-attention, temporal self-attention, and transformation attention mechanisms to create a bridge between effective spatial information and future temporal evolution information. Fourthly, we show that the STNN model can reconstruct the phase space of the dynamical system, which is explored in the time-series prediction. The experimental results demonstrate that the STNN significantly outperforms the existing methods on various benchmarks and real-world systems in the multi-step-ahead prediction of a short-term time-series.     

基于改进3D卷积网络的人体动作识别

为解决现有多数视频人体动作识别3D卷积方法无法区分信息中各维度的重要和非重要特征问题,提出了通过门控循环单元(GatedRecurrentUnit,GRU)和空间注意力增强模块构建时空特征处理网络的方法,基于多级特征融合和多组通道注意力特征选择构建网络,改进基础网络模型Res Net3D对视频人体动作识别中的网络模型.改进后模型在2个公开数据集UCF101和HMDB51上的准确率分别为96.42%和71.08%,与C3D、Two-stream等网络模型相比,具有更高的识别准确率.     

虚拟人一种导航方法研究

基于Gibson的可供性理论，提出了虚拟人一种导航方法．某个可供性是环境的不变量，是关于空间、时间和行动之间的一种联系．虚拟人能够直接感知到可供性，可以在虚拟环境中的特定区域设定语义信息，从而建立导航区域的概念．场景的八叉树用来模拟虚拟人对空间的知觉，并引入注意机制来增强感知的真实性．最后，在微机上实现了一个基于本文模型的虚拟人行为动画系统．     

Modern Science: A Case of Collective Intelligence? On the Role of Thought Economy and Gratifying Attention in Knowledge Production

Your attention please: Phenomenal conciousness, that is, how something feels, does not exist for an observer. As science relies on observations, it is not aware of the nature of subjectivity and thus science is not often defined as a collective intelligence. In this Essay, the roles of intelligence and attention are discussed, as well as an analysis of scientific communication and citation, in order to evaluate whether science is a case of collective intelligence.