基于覆盖算法的大气质量预测

提出了一种应用人工神经网络进行大气质量预测的方法,即采用多层前向网络的覆盖算法和时问序列进行短期的空气质量预测.针对空气污染的特点,选取了对空气质量有重要影响的气象因子的相关数据,并对其进行测试.实验结果表明:将该方法应用于空气质量预测,效果良好,学习速度快,识别率高,具有较强的实用价值,为实现大气质量预测提供了一种准确高效的方法.     

基于覆盖算法的大气质量预测

提出了一种应用人工神经网络进行大气质量预测的方法,即采用多层前向网络的覆盖算法和时间序列进行短期的空气质量预测。针对空气污染的特点,选取了对空气质量有重要影响的气象因子的相关数据,并对其进行测试。实验结果表明：将该方法应用于空气质量预测,效果良好,学习速度快,识别率高,具有较强的实用价值,为实现大气质量预测提供了一种准确高效的方法。     

基于深度学习的单目深度估计方法综述

深度估计是一种从单张或者多张图像预测场景深度信息的技术,是计算机视觉领域非常热门的研究方向,在三维重建、场景理解、环境感知等任务中起到了关键作用.当前深度估计技术可以分为多目深度估计和单目深度估计.因为单目摄像头具有成本低、设备较普及、图像获取方便等优点,与多目深度估计技术相比,从单目图像估计深度信息是当前更为热门和更具挑战的技术.近年来,随着深度学习的迅速发展,基于深度学习的单目深度估计方法被广泛研究.本文对基于深度估计的单目深度估计方法进行综述,首先给出单目深度估计问题的定义、介绍常用于训练的数据集与模型评价指标,然后根据不同的训练方式对国内外相关技术进行分析总结,将现有方法分为基于监督学习、无监督学习和半监督学习三大类,对每种类型方法的产生思路、优缺点进行详细分析,最后梳理、总结该技术的发展趋势与关键技术.     

深度学习在目标视觉检测中的应用进展与展望

目标视觉检测是计算机视觉领域的一个重要问题,在视频监控、自主驾驶、人机交互等方面具有重要的研究意义和应用价值.近年来,深度学习在图像分类研究中取得了突破性进展,也带动着目标视觉检测取得突飞猛进的发展.本文综述了深度学习在目标视觉检测中的应用进展与展望.首先对目标视觉检测的基本流程进行总结,并介绍了目标视觉检测研究常用的公共数据集;然后重点介绍了目前发展迅猛的深度学习方法在目标视觉检测中的最新应用进展;最后讨论了深度学习方法应用于目标视觉检测时存在的困难和挑战,并对今后的发展趋势进行展望.     

基于多尺度时空图神经网络的污染物浓度预测

基于深度学习的细粒度污染物浓度预测是一种新兴且具有前景的方法,如何充分利用气象、空间和时间等3大信息是其关键.为了协同融合3大信息,提出一种基于多尺度时空图神经网络的污染物浓度预测模型.该模型利用空气质量模型动态构建多尺度的时空图神经网络,学习污染物之间的动态时空关系.具体为:利用图神经网络学习污染物之间的多尺度空间关系,采用空气质量模型HYSPLIT构建图的结点和边属性,通过基于注意力机制的GRU (gate recurrent unit)学习污染物浓度之间的时序关系.该模型不仅充分考虑了气象、空间和时间3大影响因素,还将3个因素联动起来统一到一个框架内协同学习.该方法与传统的机理模型方法相比具有灵活部署、易于实施的特点.实际项目数据集和公开数据集上的实验表明:与现有先进的基于图神经网络的方法相比,该方法预测的污染物浓度平均绝对误差降低了0.6左右,对称平均绝对百分比误差降低0.005左右.     

机器学习在股票预测中的应用综述

揭示股票市场运行规律一直是研究的热点，近些年机器学习方法在股票预测方面取得了不错的进展，相较于传统的基本面分析、技术分析等方法，显示了独特的优势。从股票预测研究的主要问题、特征工程和机器学习算法应用等三个方面，对近年来该领域的主要文献进行总结，并针对每种算法在应用中的特点与不足进行评述。围绕目前机器学习在股票预测上遇到的主要问题，从迁移学习、特征工程、深度学习模型融合等方面进行了深入的分析与展望。     

基于音视频信息的深度多模态抑郁症识别综述

抑郁症是一种精神疾病,严重时会导致自杀行为的发生。当前抑郁症患者人数正变得越来越多,越来越普遍化、年轻化。采用机器学习方法开展面向音频、视频等模态信息的多模态抑郁症识别研究已成为一个计算机科学、心理学、医学等多学科交叉的热点课题。近年来,新发展起来的深度学习技术也逐渐被应用于面向音频、视频等模态信息的多模态抑郁症识别中的深度特征提取任务。为了系统总结和归纳近年来深度学习技术在多模态抑郁症识别领域的研究进展,首先介绍了抑郁症的临床表现及心理学诊断方法,随后简要总结了现有的抑郁症数据集,并阐述了代表性深度学习技术的基本原理及进展情况;然后,系统分析和总结了面向音频、视频的多模态抑郁症识别涉及到的关键技术,包括手工特征提取和深度特征提取,以及多模态信息融合策略;最后,指出了该领域存在的机遇与挑战,并对下一步的研究方向进行了总结与展望。     

基于注意力机制和高斯概率估计的PM2.5浓度预测

作为衡量空气污染物浓度的重要指标, 对PM_2.5浓度进行监控预测, 能够有效地保护大气环境, 进一步地减少空气污染带来的危害. 随着空气质量自动监测站的大范围建立, 由传统的机器学习搭建的空气质量预测模型已经不能满足当今的需求. 本文提出了一种基于多头注意力机制和高斯概率估计的高斯-注意力预测模型, 并对沈阳市某监测站点的数据进行了训练和测试. 该模型考虑了PM_2.5浓度受到其他空气质量数据的影响, 将空气质量数据的分层时间戳(周、日、小时)的信息对齐作为输入, 使用多头注意力机制对于不同子空间的时间序列关联特征进行提取, 能够获得更加完善有效的特征信息, 再经过高斯似然估计得到预测结果. 通过与多种基准模型进行对比, 相较于性能较优的DeepAR, 高斯-注意力预测模型的MSE、MAE分别下降了21%、15%, 有效地提高了预测准确率, 能够较准确地预测出PM_2.5浓度.     

基于离群点检测和误差修正的空气质量指数预测

空气质量指数(Air Quality Index, AQI)预测可以为人们日常生产活动以及空气污染治理工作提供指导.针对空气质量指数预测模型受离群点影响较大的问题,利用孤立森林算法对空气质量数据集进行离群点分析,采用离群鲁棒极限学习机模型(ORELM)对空气质量指数进行预测,并构建误差修正模块对模型预测误差进行修正.最后,以北京市空气质量数据作为研究对象,分别利用ORELM模型以及极限学习机(ELM)模型进行预测,并对ORELM模型预测结果进行误差修正.实验结果表明:离群鲁棒极限学习机对离群点数据集泛化性能更强,误差修正模块能有效提高模型的预测精度.     

Elman神经网络在优化空气预报模式结果中的应用

空气质量与人们的生活息息相关, 空气质量的预测结果是进行空气质量控制的依据. 因此, 提高空气质量的预测精度是本文研究的重点. CMAQ (Community Multiscale Air Quality modeling system)和CAMx (Comprehensive Air quality Model with extensions)是两种常用的空气质量数值模式, 其工作原理是通过大气物理化学方法模拟污染物传输转化过程, 进而预测空气质量. 空气质量数值模式的输入文件质量会影响到空气质量的预测精度, 为了提高空气质量预测的准确率, 本文提出了一种基于Elman神经网络的优化方法, 该方法在CMAQ和CAMx两种空气质量数值模式基础上利用Elman神经网络优化预测结果. 首先, 运行空气质量模式CMAQ和CAMx得到预测结果, 然后对预测结果进行预处理, 处理后的预测数据和实测数据一起作为Elman神经网络的输入, 进行模型的训练, 最后得到神经网络模型. 通过对测试数据集的验证和分析, 实验结果表明, 该方法表现出比单一空气质量数值模式更高的准确率.     

Multi-Site Air Pollutant Prediction Using Long Short Term Memory

The current pandemic highlights the significance and impact of air pollution on individuals. When it comes to climate sustainability, air pollution is a major challenge. Because of the distinctive nature, unpredictability, and great changeability in the reality of toxins and particulates, detecting air quality is a puzzling task. Simultaneously, the ability to predict or classify and monitor air quality is becoming increasingly important, particularly in urban areas, due to the well documented negative impact of air pollution on resident’s health and the environment. To better comprehend the current condition of air quality, this research proposes predicting air pollution levels from real-time data. This study proposes the use of deep learning techniques to forecast air pollution levels. Layers, activation functions, and a number of epochs were used to create the suggested Long Short-Term Memory (LSTM) network based neural layer design. The use of proposed Deep Learning as a structure for high-accuracy air quality prediction is investigated in this research and obtained better accuracy of nearly 82% compared to earlier records. Determining the Air Quality Index (AQI) and danger levels would assist the government in finding appropriate ways to authorize approaches to reduce pollutants and keep inhabitants informed about the findings.     

An overview of air quality analysis by big data techniques: Monitoring,forecasting, and traceability

With the rapid development of economy and the frequent occurrence of air pollution incidents, the problem of air pollution has become a hot issue of concern to the whole people. The air quality big data is generally characterized by multi-source heterogeneity, dynamic mutability, and spatial–temporal correlation, which usually uses big data technology for air quality analysis after data fusion. In recent years, various models and algorithms using big data techniques have been proposed. To summarize these methodologies of air quality study, in this paper, we first classify air quality monitoring by big data techniques into three categories, consisting of the spatial model, temporal model and spatial–temporal model. Second, we summarize the typical methods by big data techniques that are needed in air quality forecasting into three folds, which are statistical forecasting model, deep neural network model, and hybrid model, presenting representative scenarios in some folds. Third, we analyze and compare some representative air pollution traceability methods in detail, classifying them into two categories: traditional model combined with big data techniques and data-driven model. Finally, we provide an outlook on the future of air quality analysis with some promising and challenging ideas.     

Dynamic Ensemble Multivariate Time Series Forecasting Model for PM2.5

In forecasting real time environmental factors, large data is needed to analyse the pattern behind the data values. Air pollution is a major threat towards developing countries and it is proliferating every year. Many methods in time series prediction and deep learning models to estimate the severity of air pollution. Each independent variable contributing towards pollution is necessary to analyse the trend behind the air pollution in that particular locality. This approach selects multivariate time series and coalesce a real time updatable autoregressive model to forecast Particulate matter (PM) PM2.5. To perform experimental analysis the data from the Central Pollution Control Board (CPCB) is used. Prediction is carried out for Chennai with seven locations and estimated PM’s using the weighted ensemble method. Proposed method for air pollution prediction unveiled effective and moored performance in long term prediction. Dynamic budge with high weighted k-models are used simultaneously and devising an ensemble helps to achieve stable forecasting. Computational time of ensemble decreases with parallel processing in each sub model. Weighted ensemble model shows high performance in long term prediction when compared to the traditional time series models like Vector Auto-Regression (VAR), Autoregressive Integrated with Moving Average (ARIMA), Autoregressive Moving Average with Extended terms (ARMEX). Evaluation metrics like Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and the time to achieve the time series are compared.     

Air pollution forecasting based on attention-based LSTM neural network and ensemble learning

With air pollution having become a global concern, scientists are committed to working on its amelioration. In the field of air pollution prediction, there have been good results in experimental research so far, but few studies have integrated weather forecast information and the properties of air pollution drift. In this work, we propose a novel wind-sensitive attention mechanism with a long short-term memory (LSTM) neural network model to predict the air pollution - PM2.5 concentrations by considering the influence of wind direction and speed on the changes of spatial–temporal PM2.5 concentrations in neighbouring areas. Preliminary predictions for PM2.5 are then made by an LSTM neural network regarding neighbouring pollution; these predictions are “paid attention to” and we finally apply an ensemble learning method based on e X treme G radient B oosting (XGBoost) to combine the preliminary predictions with weather forecasting to make second phase predictions of PM2.5. The experiment is conducted using PM2.5 data and weather forecast data. Our results illustrate that the proposed method is superior to other methods in predicting PM2.5 concentrations, including multi-layer perceptron, support vector regression, LSTM neural network, and extreme gradient boosting algorithm.     

深度学习认知计算综述

随着大数据和智能时代的到来,机器学习的研究重心已开始从感知领域转移到认知计算（Cognitive computing,CC）领域,如何提升对大规模数据的认知能力已成为智能科学与技术的一大研究热点,最近的深度学习有望开启大数据认知计算领域的研究新热潮.本文总结了近年来大数据环境下基于深度学习的认知计算研究进展,分别从深度学习数据表示、认知模型、深度学习并行计算及其应用等方面进行了前沿概况、比较和分析,对面向大数据的深度学习认知计算的挑战和发展趋势进行了总结、思考与展望.     

Empirical analysis of session-based recommendation algorithms

Recommender systems are tools that support online users by pointing them to potential items of interest in situations of information overload. In recent years, the class of session-based recommendation algorithms received more attention in the research literature. These algorithms base their recommendations solely on the observed interactions with the user in an ongoing session and do not require the existence of long-term preference profiles. Most recently, a number of deep learning-based (“neural”) approaches to session-based recommendations have been proposed. However, previous research indicates that today’s complex neural recommendation methods are not always better than comparably simple algorithms in terms of prediction accuracy. With this work, our goal is to shed light on the state of the art in the area of session-based recommendation and on the progress that is made with neural approaches. For this purpose, we compare twelve algorithmic approaches, among them six recent neural methods, under identical conditions on various datasets. We find that the progress in terms of prediction accuracy that is achieved with neural methods is still limited. In most cases, our experiments show that simple heuristic methods based on nearest-neighbors schemes are preferable over conceptually and computationally more complex methods. Observations from a user study furthermore indicate that recommendations based on heuristic methods were also well accepted by the study participants. To support future progress and reproducibility in this area, we publicly share the session-rec evaluation framework that was used in our research.

     

基于深度学习的流行度预测研究综述

在线社交网络中的消息流行度预测研究,对推荐、广告、检索等应用场景都具有非常重要的作用.近年来,深度学习的蓬勃发展和消息传播数据的积累,为基于深度学习的流行度预测研究提供了坚实的发展基础.现有的流行度预测研究综述,主要是围绕传统的流行度预测方法展开的,而基于深度学习的流行度预测方法目前仍未得到系统性地归纳和梳理,不利于流...     

基于多元回归分析的PM2.5预测研究

PM2.5对人体健康和大气环境质量的影响众所周知,分析、预测PM2.5浓度对污染天气防治与干预有着非常重要的作用。利用灰色关联度、多元回归分析等方法对全国各大城市空气质量进行了研究,分析了影响PM2.5浓度的主要因素并进行了影响程度排序,构建了PM2.5预测模型并进行了预测实践,为我国环境空气质量预报和污染天气防治干预提供了有效的决策信息。     

深度学习实体关系抽取研究综述

实体关系抽取作为信息抽取、自然语言理解、信息检索等领域的核心任务和重要环节,能够从文本中抽取实体对间的语义关系.近年来,深度学习在联合学习、远程监督等方面上的应用,使关系抽取任务取得了较为丰富的研究成果.目前,基于深度学习的实体关系抽取技术,在特征提取的深度和模型的精确度上已经逐渐超过了传统基于特征和核函数的方法.围绕有监督和远程监督两个领域,系统总结了近几年来中外学者基于深度学习的实体关系抽取研究进展,并对未来可能的研究方向进行了探讨和展望.