Synaesthesia and cycling data art: towards cross-modal representations of self-tacking cycling data

This paper develops ideas around cycling as art practice. The questions at the heart of it revolve around how new data technologies enable us to represent such experiences as artworks. Recent research in neuroscience has begun to establish the ways in which modes of perception are processed in the brain. Related research suggests that Synaesthesia may be caused by a genetic mutation that results in a ‘cross wiring’ of these modes of perception. Within the arts there is a long history of exploration around synaesthesia, ideas that are becoming relevant again particularly in relation to the growth in personal data. Representing sensations and transferring them from one mode to another offers a way to handle some of this data and potentially say new things about our experiences. The author explores these ideas through artworks made using data from cycling to investigate the visualization of this experience.     

What is generative art?

There are various forms of what's sometimes called generative art, or computer art. This paper distinguishes the major categories and asks whether the appropriate aesthetic criteria—and the locus of creativity—are the same in each case.     

Deep Teleology in Artificial Systems

Teleological variations of non-deterministic processes are defined. The immediate past of a system defines the state from which the ordinary (non-teleological) dynamical law governing the system derives different possible present states. For every possible present state, again a number of possible states for the next time step can be defined, and so on. After k time steps, a selection criterion is applied. The present state leading to the selected state after k time steps is taken to be the effective present state. Hence, the present state of a system is defined by its past in the sense that the past determines the possible states that are to be considered, and by its future in the sense that the selection of a possible future state determines the effective present state. A system that obeys this type of teleological dynamics may have significantly better performance than its non-teleological counterpart. The basic reason is that evolutions that are less optimal for the present time step, but which lead to a higher optimality after k time steps, may be preferred. This abstract concept of teleology is implemented for two concrete systems. First, it is applied to a general method for function approximation and classification problems. The method at issue treats all problems handled by conventional connectionism, and is suited for information with inner structure also. Second, it is applied to a dynamics in which forms of maximal homogeneity have to be produced. The relevance of the latter dynamics for generative art is illustrated. The teleology is `deep' in the sense that it is situated at the cellular level, in contradistinction with the teleology that is usually met in cognitive contexts, and which refers to macroscopic processes such as making plans. It is conjectured that deep level teleology is useful for machines, even though the issue if natural systems use this teleology is left open.     

Personal informatics for everyday life: How users without prior self-tracking experience engage with personal data

The spreading of devices and applications allowing people to collect personal information opens new opportunities for Personal Informatics. Although many of these tools are already effectively used by motivated people to gain self-knowledge and produce change in their behaviors, there is a great number of users that are potentially interested in Personal Informatics but do not know of its potentialities and criticalities. In order to investigate how users perceive and use self-tracking tools in everyday life, we conducted a diary study, requiring fourteen participants with no previous experience in Personal Informatics to use a variety of trackers. We discovered that they use and perceive these technologies differently from the ones experienced in self-tracking. Participants considered the act of collecting personal information burdensome, with no beneficial reward. We also uncovered a series of problems that they experienced while tracking, managing, visualizing, and using their data. Among them we found that the lack of suggestions on using data and the excess of abstract visualization in the apps prevented users to gain useful insights. As a result, their interest in self-tracking soon faded, despite their initial curiosity in exploring and “playing” with their data. Starting from the findings of this study, we identified seven design strategies to better Personal Informatics tools, supported by literature and examples that draw from different research fields, from tangible interfaces, to virtual environments and video games. These strategies are primarily addressed to satisfy the inexperienced users' needs, but their applicability can be reasonably extended to all the individuals curious and interested in Personal Informatics.     

Understanding persistence in the use of Online Fitness Communities: Comparing novice and experienced users

Mobile and wearable technologies facilitate physiological data collection for health and wellness purposes. Users typically access these data via Online Fitness Community (OFC) platforms (e.g., Fitbit, Strava, RunKeeper). These platforms present users with functionalities centered on self-monitoring, social networking and enjoyment. In order to fully benefit from these functionalities, users need to make a habit out of integrating OFC use into their everyday workout routines. However, research suggests that users often fail to use OFCs over a longer period of time. This study sheds light on the factors that explain persisted OFC use. To that end, the study compares novice and experienced users in terms of their OFC use motives and how these motives contribute to the habitual integration of OFCs into everyday workout routines. Based on the survey responses of 394 OFC users, a multi-sample structural equation model indicates that self-regulatory and social motives directly predict habitual OFC use, and that enjoyment and self-regulatory motives indirectly predict habitual OFC use, by driving the perceived usefulness of OFCs. Moderation analysis revealed that, for novice users, self-regulatory motives are the prime drivers of habitual OFC use, while social motives and enjoyment are more important for experienced users.     

Exploring creative intersections: Ernest Edmonds and his time-based generative art

This article focuses on the experience of British artist Ernest Edmonds and the influences that have informed his art practice in the past forty years. How are these influences connected? The article has an historical focus. It develops the themes within Edmonds's art and shows his connections with the ‘Systems’ artists and their forebears. In particular, the article concentrates on the encounter of Edmonds with artistic thinking about systems and process in the broad sense, as well as digital and interactive work developed from the early 1980s. As the article will demonstrate, a passion for colour, time and structure, the encounter with a number of artists inspired by Constructivism in the early 1980s, and the educational context in which Edmonds has worked from the late 1960s until the present, have offered great opportunities of interdisciplinary exchanges and ideas that had a profound impact upon the nature of his art. These influences have enabled Edmonds to explore new constructs in art through the use of technology that are a constant stimulus in his creative research, both as an artist and as an academic.     

基于生成艺术的空间形态生成研究

有着2500年历史的绍兴市,以纵横交错的河道、众多的石桥和乌篷船而闻名于世。本文基于生成艺术的思想,采用GIS技术和三维建模技术生成了一种绍兴市独有的空间形态（乌篷形态）,该形态对于计算机创作和建筑设计领域具有积极的启发意义。     

VALENCE: affective visualisation using EEG

VALENCE is an interactive visualisation controlled by live brainwave monitoring. We used a wireless EEG headset to monitor the player's alpha waves (an indicator of relaxation) and valence (an indicator of emotion or arousal). The game world is an emergent system of attractive and repulsive forces responding to EEG input.     

Relatively quantified constraint satisfaction

The constraint satisfaction problem (CSP) is a convenient framework for modelling search problems; the CSP involves deciding, given a set of constraints on variables, whether or not there is an assignment to the variables satisfying all of the constraints. This paper is concerned with the more general framework of quantified constraint satisfaction, in which variables can be quantified both universally and existentially. We study the relatively quantified constraint satisfaction problem (RQCSP), in which the values for each individual variable can be arbitrarily restricted. We give a complete complexity classification of the cases of the RQCSP where the types of constraints that may appear are specified by a constraint language.     

Cancer classification with data augmentation based on generative adversarial networks

Accurate diagnosis is a significant step in cancer treatment. Machine learning can support doctors in prognosis decision-making, and its performance is always weakened by the high dimension and small quantity of genetic data. Fortunately, deep learning can effectively process the high dimensional data with growing. However, the problem of inadequate data remains unsolved and has lowered the performance of deep learning. To end it, we propose a generative adversarial model that uses non target cancer data to help target generator training. We use the reconstruction loss to further stabilize model training and improve the quality of generated samples. We also present a cancer classification model to optimize classification performance. Experimental results prove that mean absolute error of cancer gene made by our model is 19.3% lower than DC-GAN, and the classification accuracy rate of our produced data is higher than the data created by GAN. As for the classification model, the classification accuracy of our model reaches 92.6%, which is 7.6% higher than the model without any generated data.     

Declarative generation of synthetic XML data

Synthetic data can be extremely useful in testing and evaluating algorithms, tools and systems. Most synthetic data generators available today are the result of individual benchmarking efforts. Typically, these are complex programs in which the specifications of both the structure and the contents of the data are hard‐coded. As a result, it is often difficult to customize these tools for producing synthetic data tailored for specific needs. In this article, we describe the ToXgene synthetic data generator, which is a declarative tool for generating realistic XML data for benchmarking as well as testing purposes. We present our template specification language, which consists of augmenting XML Schema with probabilistic models that guide the data‐generation process. We discuss the architecture of our current implementation and we argue about ToXgene's usefulness by discussing experimental results as well as describing two projects that use our tool. Copyright © 2006 John Wiley & Sons, Ltd.     

面向不平衡网络评论数据挖掘的服务质量评价

针对基于不均衡网络评论的服务质量评价研究中存在的评论语料分布不均衡和特征词提取依赖主观经验的问题,在利用不平衡网络评论构建服务质量评价指标体系的过程中,本文提出基于主体频率G指数模型来客观地提取高频特征词,并通过特征词词向量聚类构建服务质量评价指标体系.在服务质量评分模型中,本文提出一种基于热度指数加权的顾客满意度评分...     

基于多尺度生成对抗网络的遮挡行人重识别方法

行人重识别是指在多个非重叠摄像头拍摄的场景下,给定一幅查询行人图像,从大规模行人图像库中检索出具有相同身份的行人图像,是一类特殊的图像检索任务.随着深度学习的不断发展,行人重识别方法的性能得到了显著提升.但是行人重识别在实际应用中经常遭遇遮挡问题(例如,背景遮挡、行人互相遮挡等).由于遮挡图像不仅丢失了部分目标信息,而且引入了额外的干扰,使得现有方法往往难以学习到鲁棒的特征表示,从而导致识别性能严重下降.最近,生成对抗网络在各类计算机视觉任务上展现出强大的图像生成能力.受到生成对抗网络的启发,提出了一种基于多尺度生成对抗网络的遮挡行人重识别方法.首先,利用成对的遮挡图像和非遮挡图像训练一个多尺度生成器和一个判别器.多尺度生成器能够对随机遮挡区域进行去遮挡操作,生成高质量的重构图;而判别器能够区分输入图像是真实图像还是生成图像.其次,利用训练好的多尺度生成器,生成去除随机遮挡的训练图像,添加到原始训练图像集,用于增加训练样本的多样性.最后,基于此扩充训练图像集,训练分类识别模型,有效地提高模型在测试图像集上的泛化性.在多个有挑战性的行人重识别数据集上的实验结果,验证了所提出方法的有效性.     

基于概率生成模型的网络数据分类方法

利用实体之间的相互关系来对实体进行分类的网络数据分类是数据挖掘的一个重要研究内容.现有的网络数据分类方法普遍根据邻居节点的类别来对节点进行分类.这些方法在同质性程度较高的网络中达到了很高的分类精度.然而在现实世界中,存在许多同质性程度很低的网络.在低同质性网络中,大多数相连节点的类别不同,所以现有方法难以正确预测出节点的类别.因此,提出了一种新的网络数据分类方法.其主要思路是建立一个描述网络的概率生成模型.在这个概率生成模型中,将网络中的边作为观察变量,将未知类别节点的类别作为潜在变量.通过吉布斯采样方法对模型进行求解,计算出潜在变量的取值,从而得到未知类别节点的类别.在真实数据集上的对比实验表明,提出的分类方法在低同质性网络上有更好的分类性能.     

支持差分隐私的图像数据挖掘方法研究

针对数据挖掘模型中存在的隐私泄漏问题及现有隐私保护技术的不透明性,本文将差分隐私与图像生成模型生成对抗网络(Generative adversarial network,GAN)相结合,提出了一种更具普适性的支持图像数据差分隐私保护的生成对抗网络模型(Image differential privacy-GAN,IDP...     

基于生成对抗网络和变分自编码器的离群点检测算法

针对传统离群点检测算法在类极度不平衡的高维数据集中难以学习离群点的分布模式,导致检测率低的问题,提出了一种生成对抗网络(generative adversarial network,GAN)与变分自编码器(variational auto-encoder,VAE)结合的GAN-VAE算法。算法首先将离群点输入VAE训练,学习离群点的分布模式;然后将VAE与GAN结合训练,生成更多潜在离群点,同时学习正常点与离群点的分类边界;最后将测试数据输入训练后的GAN-VAE,根据正常点与离群点相对密度的差异性计算每个对象的离群值,将离群值高的对象判定为离群点。在四个真实数据集上与六个离群点检测算法进行对比实验,结果表明GAN-VAE在AUC、准确率和F;值上平均提高了5.64%、5.99%和13.30%,证明GAN-VAE算法是有效可行的。     

GAN-TM:一种基于生成对抗网络的流量矩阵推断机制

随着当前互联网技术的快速发展,网络规模和复杂度不断提高,由于流量矩阵对于网络管理、流量工程、异常检测等都具有重要意义,因此准确测量流量矩阵对于计算机网络而言极其重要.当前针对流量矩阵的测量机制主要可以分为直接测量法和估计推断法,其中估计方法又包括简单统计反演法、附加链路测量信息法以及测量反演结合法.现有测量机制在准确性...     

Low-level dichotomy for quantified constraint satisfaction problems

Building on a result of Larose and Tesson for constraint satisfaction problems (CSPs), we uncover a dichotomy for the quantified constraint satisfaction problem QCSP(B), where B is a finite structure that is a core. Specifically, such problems are either in ALogtime or are L-hard. This involves demonstrating that if CSP(B) is first-order expressible, and B is a core, then QCSP(B) is in ALogtime.We show that the class of B such that CSP(B) is first-order expressible (indeed trivial) is a microcosm for all QCSPs. Specifically, for any B there exists a C — generally not a core — such that CSP(C) is trivial, yet QCSP(B) and QCSP(C) are equivalent under logspace reductions.     

Value ordering for quantified CSPs

We investigate the use of value ordering in backtracking search for Quantified Constraint Satisfaction problems (QCSPs). We consider two approaches for ordering heuristics. The first approach is solution-focused and is inspired by adversarial search: on existential variables we prefer values that maximise the chances of leading to a solution, while on universal variables we prefer values that minimise those chances. The second approach is verification-focused, where we prefer values that are easier to verify whether or not they lead to a solution. In particular, we give instantiations of this approach using QCSP-Solve’s pure-value rule Gent et al. (QCSP-solve: A solver for quantified constraint satisfaction problems. In Proceedings of IJCAI, pp. 138–143, 2005). We show that on dense 3-block problems, using QCSP-Solve, the solution-focused adversarial heuristics are up to 50% faster than lexicographic ordering, while on sparse loose interleaved problems, the verification-focused pure-value heuristics are up to 50% faster. Both types are up to 50% faster on dense interleaved problems, with one pure-value heuristic approaching an order of magnitude improvement. This work was supported in part by Microsoft Research through the European PhD Scholarship Programme, and by the Embark Initiative of the Irish Research Council for Science, Engineering and Technology (IRCSET).