Notions of explainability and evaluation approaches for explainable artificial intelligence

Explainable Artificial Intelligence (XAI) has experienced a significant growth over the last few years. This is due to the widespread application of machine learning, particularly deep learning, that has led to the development of highly accurate models that lack explainability and interpretability. A plethora of methods to tackle this problem have been proposed, developed and tested, coupled with several studies attempting to define the concept of explainability and its evaluation. This systematic review contributes to the body of knowledge by clustering all the scientific studies via a hierarchical system that classifies theories and notions related to the concept of explainability and the evaluation approaches for XAI methods. The structure of this hierarchy builds on top of an exhaustive analysis of existing taxonomies and peer-reviewed scientific material. Findings suggest that scholars have identified numerous notions and requirements that an explanation should meet in order to be easily understandable by end-users and to provide actionable information that can inform decision making. They have also suggested various approaches to assess to what degree machine-generated explanations meet these demands. Overall, these approaches can be clustered into human-centred evaluations and evaluations with more objective metrics. However, despite the vast body of knowledge developed around the concept of explainability, there is not a general consensus among scholars on how an explanation should be defined, and how its validity and reliability assessed. Eventually, this review concludes by critically discussing these gaps and limitations, and it defines future research directions with explainability as the starting component of any artificial intelligent system.     

A Brief Review of Artificial Intelligence Applications and Algorithms for Psychiatric Disorders

A number of brain research projects have recently been carried out to study the etiology and mechanisms of psychiatric disorders. Although psychiatric disorders are part of the brain sciences, psychiatrists still diagnose them based on subjective experience rather than by gaining insights into the pathophysiology of the diseases. Therefore, it is urgent to have a clear understanding of the etiology and pathogenesis of major psychiatric diseases, which can help in the development of effective treatments and interventions. Artificial intelligence (AI)-based applications are being quickly developed for psychiatric research and diagnosis, but there is no systematic review that summarizes their applications. For this reason, this study briefly reviews three main brain observation techniques used to study psychiatric disorders—namely, magnetic resonance imaging (MRI), electroencephalography (EEG), and kinesics diagnoses—along with related AI applications and algorithms. Finally, we discuss the challenges, opportunities, and future study directions of AI-based applications.     

基于时序数据的云网协同平台人工智能运维体系

云计算在企业应用中的拓展不但表现为平台规模的拓展，也表现为平台应用的延伸。“云网协同”和“微服务化”是当前企业云平台演进的重要趋势。随着企业信息化建设重要性的持续提升，微服务化云网协同平台的运行维护面临极大挑战。首先分析了平台运维面临的挑战，梳理了平台人工智能运维需求，提出了基于时序数据分析的平台人工智能运维技术体系，并给出了云网协同平台人工智能运维子系统参考设计。所提技术体系和系统设计具有实用性和推广性，可以作为企业云平台建设和优化的技术途径参考。     

水泥工厂视频监控智能预警平台研究及规划

视频监控系统作为智能化工厂体系架构的子模块，在水泥工厂得到了广泛的部署和应用，为安全、生产流程等管理需求提供了远程可视化的技术支撑手段。为进一步提升工厂智能化水平，提高管理效率，本文提出了基于视频图像分析技术的智能监控预警平台的路线规划，实施步骤和方法，并对该平台的初步应用效果进行了量化评估，为后续平台建设提供科学决策依据。     

Factors influencing the use of artificial intelligence in government: Evidence from China

Some studies have discussed the potential and challenges related to the use of artificial intelligence (AI) in government. However, there are few empirical studies that have examined factors that influence the use of AI in government. By collecting policy documents and empirical data from the government, IT enterprises, and the public in China, we identified the influencing factors in the three stages of government adoption, implementation, and decision-making. The research results show that the influencing factors of government application of AI are different at different stages and with different stakeholders’ backgrounds.     

Performance improvement of microbial fuel cell through artificial intelligence

The current work introduces an enhancement in the performance of the microbial fuel cell through estimating the optimal set of controlling parameters. The maximization of both power density (PD) and the percentage of chemical oxygen demand (COD) removal were considered as the enhancement in the cell's performance. Three main parameters in terms of performance as well as commercialization are the system's inputs; the Pt which takes the range of 0.1‐0.5 mg/cm², the degree of sulphonation in sulfonated‐poly‐ether‐ether‐ketone that changes in the range of 20‐80%, and the rate of aeration of cathode which varies between 10 and 150 mL/min. From the experimental dataset, two robust adaptive neuro‐fuzzy inference system models based on the fuzzy logic technique have been constructed. The comparisons between the models' outputs and the experimental data showed well‐fitting in both training and testing datasets. The mean squared errors of the PD model, for testing and whole datasets, were found 2.575 and 0.909 while for the COD model it showed 19.242 and 6.791, respectively. Then, based on the two fuzzy models, a Particle Swarm Optimization algorithm has been used to determine the best parameters that maximize both of the PD and the COD removal of the cell. The optimization process was utilized for single and multi‐object optimization processes. In the single optimization, the resulting maximums of the PD and the COD removal were found 62.844 (mW/m²) and 99.99 (%), respectively. Whereas, in the multi‐object optimization, the values of 61.787 (mW/m²) and 96.21 (%) were reached as the maximums for the PD and COD, respectively. This implies that, in both cases of optimization processes, the adopted methodology can efficiently enhance the microbial fuel cell performances than the previous work.     

个性化学习路径推荐综述

近年来，伴随着现代信息技术的迅猛发展，以人工智能为代表的新兴技术在教育领域得到了广泛应用，引发了学习理念和方式的深刻变革.在这种大背景下，在线学习超越了时空的限制，为学习者“随时随地”学习提供了更多的可能性，从而得到了蓬勃发展.然而，在线学习中师生时间、空间分离的特征，导致教师无法及时掌握学生的学习状态，一定程度上制约了在线学习中教学质量的提升.面对多元化的学习需求及海量学习资源，如何迅速完成学习目标、降低学习成本、合理分配学习资源等问题成为限制个人和时代发展的重大问题.然而，传统的“一刀切”的教育模式已经不能满足人们获取知识的需求了，需要一个更高效、更科学的个性化教育模式，以帮助学习者以最小的学习成本最大限度地完成学习目标.基于以上背景，如何自动高效识别学习者特征，高效地组织和分配学习资源，为每一位学习者规划个性化路径，成为面向个体的精准化教育资源匹配机制研究中亟待解决的问题.系统地综述并分析了当前个性化学习路径推荐的研究现状，并从多学科领域的角度分析了对于同一问题的不同研究思路，同时也归纳总结了当前研究中最为主流的核心推荐算法.最后，强调当前研究存在的主要不足之处.     

竞技二打一游戏中同等牌力的研究

全国竞技二打一扑克锦标赛是国家体育总局棋牌运动管理中心推出的新兴项目，比赛采用复式赛赛制，然而这种赛制存在无法杜绝线上比赛作弊的问题，为此提出用同等牌力赛制来代替复式赛制进行比赛的设想。本文主要针对如何判定和选取具有同等牌力的初始手牌进行研究，提出用等级难度评估指标来计算初始手牌的难度，初始手牌难度相近的牌认定为同等牌力，并给出了同等牌力的选取方法；为保证一副牌所得等级难度评估指标的可靠性，提出用多个不同水平的“二打一AI程序”来代替真人打牌产生牌谱的方法。实验证明了同等牌力生成方法的有效性，本文提出的同等牌力方法应用于比赛是可行的。     

Assessing Industrial Robot agility through international competitions

Manufacturing and Industrial Robotics have reached a point where to be more useful to small and medium sized manufacturers, the systems must become more agile and must be able to adapt to changes in the environment. This paper describes the process for creating and the lessons learned over multiple years of the Agile Robotics for Industrial Automation Competition (ARIAC) being run by the National Institute of Standards and Technology.     

Modified OFS-RDS bat algorithm for IFS encoding of bitmap fractal binary images

This work is an extension of a previous paper (presented at the Cyberworlds 2019 conference) introducing a new method for fractal compression of bitmap binary images. That work is now extended and enhanced through three new valuable features: (1) the bat algorithm is replaced by an improved version based on optimal forage strategy (OFS) and random disturbance strategy (RDS); (2) the inclusion of new similarity metrics; and (3) the consideration of a variable number of contractive maps, whose value can change dynamically over the population and over the iterations. The first feature improves the search capability of the method, the second one improves the reconstruction accuracy, and the third one computes the optimal number of contractive maps automatically. This new scheme is applied to a benchmark of two binary fractal images exhibiting a complex and irregular fractal shape. The graphical and numerical results show that the method performs very well, being able to reconstruct the input images with high accuracy. It also computes the optimal number of contractive maps in a fully automatic way. A comparative work with other alternative methods described in the literature is also carried out. It shows that the presented method outperforms the previous approaches significantly.