Deep learning: emerging trends, applications and research challenges
Deep learning: emerging trends, applications and research challenges
复制标题
DOI：
10.1007/s00500-020-04939-z
复制
发表时间：
2020-04
期刊：
影响因子：
4.1
通讯作者：
Mu-Yen Chen;Hsiu-Sen Chiang;E. Lughofer;E. Eğrioğlu
中科院分区：
文献类型：
--
作者：
Mu-Yen Chen;Hsiu-Sen Chiang;E. Lughofer;E. Eğrioğlu
机器学习是设计和分析算法，使计算机能够自动“学习”，并允许机器通过自动分析数据并使用它们来预测未知数据来建立规则。传统的机器学习方法从问题定义、适当的信息收集到模型开发和结果验证的过程已经过时，难以满足物联网的需求。但是，由于人工智能（AI）和高速计算性能的发展，最近的情况发生了巨大变化。因此，深度学习是一个很好的例子，它通过特征工程打破了机器学习的限制，并提供了令人难以置信的上级性能。它使许多非常复杂的应用成为可能。机器学习已经被应用于解决人类社会中的复杂问题多年，机器学习的成功是因为计算能力以及传感技术的支持。人工智能和数据驱动方法的发展将很快对该领域产生重大影响。搜索引擎、图像识别、生物识别、语音和手写识别、自然语言处理，甚至医疗诊断和金融信用评级都是常见的例子。很明显，随着人工智能渗透到我们的世界，更具体地说，我们的生活，将给公众带来许多挑战。因此，本特刊旨在汇集各种研究和开发成果，探索深度学习背景下人工智能发展所面临的技术、应用和挑战。论文的简要概述和讨论如下：本特刊的第一个主题关注“深度学习模型中参数的减少”。Cao和Wang（2019）将主成分分析（PCA）和反向传播（BP）神经网络算法相结合，构建了股票价格预测模型。实验结果表明，该方法较传统的投资策略有明显的改善。Chen和Huang（2019）使用演化矩阵函数提取重要特征，然后构建3D艺术创作网络。Gao等人（2019）将遗传算法（GA）应用到BP神经网络中，实验结果表明该模型可以获得较高的可转换债券和封闭式基金收益的准确性。Huang等人（2019 a，B，c）运用灰色关联多目标决策方法提取原住民老年人福尔斯跌倒的高危因素。该模型可用于神经网络构造中的变量约简。Liu（2019 a，B）基于灰色关联分析（GRA）开发了新型层次分析法（AHP）。该方法可以有效地去除模型深度学习中不必要的因素和参数。Ohno（2019）使用变分自编码器（VAE）作为数据增强的生成模型，可以成为构建多层神经网络的有效方法。Sangaiah等人（2019a）使用基于地理的优化（BBO）来减少参数，实验结果说明了所提出的算法的有效性和可行性。Sangaiah等人（2019 b）开发了布谷鸟优化算法（COA）来构建鲁棒混合整数线性
Machine learning is to design and analyze algorithms that allow computers to ‘‘learn’’automatically, and allows machines to establish rules from automatically analyzing data and using them to predict unknown data. Traditional machine learning approach is difficult to meet the needs of Internet of Things (IoT) only through its outdated process starting from problem definition, appropriate information collection, and ending with model development and results verification. But however, recent scenario has dramatically changed due to the development of artificial intelligence (AI) and high-speed computing performance. Therefore, deep learning is a good example that breaks the limits of machine learning through feature engineering and gives astonishingly superior performance. It makes a number of extremely complex applications possible. Machine learning has been applied to solve complex problems in human society for years, and the success of machine learning is because of the support of computing capabilities as well as the sensing technology. An evolution of artificial intelligence and data-driven approaches will soon cause considerable impacts to the field. Search engines, image recognition, biometrics, speech and handwriting recognition, natural language processing, and even medical diagnostics and financial credit ratings are all common examples. It is clear that many challenges will be brought to publics as the artificial intelligence infiltrates into our world, and more specifically, our lives. Thus, this special issue aims to bring together various research and development achievements in exploring techniques, applications, and challenges that face the evolution of artificial intelligence in the context deep learning. A brief overview of the papers is presented and discussed as follows:The first theme in this special issue focuses on ‘‘Reduction of parameters in deep-learning models’’. Cao and Wang (2019) integrated the principal component analysis (PCA) and back propagation (BP) neural network algorithm to construct the stock price prediction model. The experimental results illustrated the significant improvement than the traditional investment strategies. Chen and Huang (2019) used the evolution matrix function to extract the important feature and then built the 3D art creation network. Gao et al.(2019) applied the genetic algorithm (GA) into the BPNN and the experimental results illustrated the model can obtain the high accuracy of the convertible bonds and closed funds benefits. Huang et al.(2019a, b, c) used the grey relational and multi-objective decisionmaking methods to extract the high risk factors for aboriginal elderly falls. The proposed model can be useful for the variables reduction in the neural network construction. Liu (2019a, b) developed the novel analytic hierarchy process (AHP) based on the grey relational analysis (GRA). The proposed method can be the efficient way to recede the unnecessary factors and parameters in deeplearning models. Ohno (2019) used the variational autoencoders (VAEs) as generative models for data augmentation and can be an efficient method to build the multilayer neural networks. Sangaiah et al.(2019a) used the biogeography-based optimization (BBO) to reduce the parameters, and experimental results illustrated the efficiency and feasibility of the proposed algorithm. Sangaiah et al.(2019b) developed the cuckoo optimization algorithm (COA) to construct the robust mixed-integer linear