Self-Organized Cognitive Maps

This paper argues cognitive mapping is a learning process that can be simulated by a self-organizing neural network. The learning of city locations was considered in two studies. One study focused on the learning of city locations on four continents. Results indicated the neural network aligned the cities producing systematic errors similar to those in human cognitive maps. A second study had a neural network learn a biased sample of city locations in the United States. Results indicated a non-linear relationship between cognitive and physical distances. Self-organized cognitive maps naturally produce this non-linear relationship when information from more than one scale is mapped into one space.     

Tasks,Strategies, and Cognitive Processes Associated With Navigational Map Reading: A Review Perspective*

All geographers use and may create maps in their professional activities. While perceived by some as merely tools to present spatial information, maps are actually complicated communication mechanisms. Many geographers spend tremendous amounts of time learning about verbal and written communication (how to speak, listen, read critically, and write effectively). But to many, one of the basic tools of geographic communication, the map, is a general source of unknown. Many previous papers have focused on how to create effective maps; this paper focuses on how people read and use navigation maps. Everyone completes map‐reading tasks, uses (often different) strategies, and processes information through cognitive processing. Geographers may benefit from understanding how they, as well as their map‐viewing public, process the information contained on their maps. This paper addresses terms and issues addressed in navigational map reading, specifically including tasks and strategies as well as the governing cognitive processes.     

EVIDENCE OF A RESPONSE-BIAS EXPLANATION OF NONEUCLIDEAN COGNITIVE MAPS*

Two groups of subjects with varying amounts of experience with the city in which they lived judged from memory distances between familiar locations. The distance judgments were processed by a multidimensional scaling algorithm (MULTISCALE) that estimates the parameters of a power function transformation of the response scale and the coordinates of the locations in k-dimensional euclidean space. The results for both subject groups suggested that inaccuracies of distance cognition do not reflect the noneuclidean nature of urban cognitive maps but are caused by response bias.     

Prediction of Anopheles minimus habitat in India—a tool for malaria management

This paper reports research to predict the distribution of An. minimus, a malaria vector in forest fringe areas using GIS to support precision surveys for malaria control. Because An. minimus is a forest‐associated species, generalized thematic maps (1:6?000?000) of forest cover, soil type, altitude, rainfall and temperature were used. Digitization, overlaying, integration and analysis of thematic maps were done using Arc/Info 8.1 NT and Arc/View 3.2 (GIS, ESRI) software. GIS delineated favourable areas for An. minimus where the species is likely to be found, and precision surveys can be conducted. Precision field surveys in selected locations of favourable/non‐favourable areas were carried out. The species could be found in all locations designated as a favourable area and was absent in non‐favourable areas. In two districts, one where the species is reported to have disappeared in the early 1950s and the other where it was not reported in earlier surveys, GIS helped in precision surveys, and An. minimus was found. The technique can quickly cover vast and inaccessible areas and is easy to duplicate in other parts of the world to assist cost‐effective control of malaria. It can also delineate areas favourable for any species of flora and fauna to help precision surveys.     

认知地图的地理学研究进展与展望

认知地图作为人对人地关系复杂系统进行抽象化和秩序化的内在表征,是个体空间行为决策、空间认知与空间规划的理论基础,也是地理学的重要研究领域之一。论文通过CiteSpace文献学分析方法并结合经典文献回顾,从多学科视角系统梳理认知地图1948—2020年共70余年的发展脉络与研究热点。研究认为,认知地图在地理学的学科发展过程中扮演了重要角色,特别是在认知空间与城市意象、空间扭曲与空间偏好、空间知识与决策行为等方面丰富了地理学内涵。近期与地理学有关的认知地图跨学科研究出现3大转向：大数据时代重新定义认知空间与城市意象的数字转向,以特定群体或事件的空间偏好为切入点的情感转向,以第一人称视角解译空间知识与决策行为的神经转向。最后指出认知地图空间分析的地理学在跨学科融合发展中的角色与应用前景。     

基于跟踪调查的北京城市空间感知研究——以中国科学院大学硕士研究生为例

城市空间感知是行为地理学的经典研究课题,论文采用认知地图并辅以问卷调查方法,在一年时间内跟踪调查中国科学院大学雁栖湖校区2017级硕士研究生,根据手绘地图分析中国科学院大学(国科大)硕士研究生城市空间感知的过程与特征。结果表明：① 有别于凯文·林奇(K. Lynch)的空间意象5大要素,国科大研究生手绘地图的基本要素主要为标志物、节点、道路,其中标志物占主导地位;② 根据要素组合特征,可将认知地图划分为散点型、过渡型和拓扑型3类,散点型地图占据绝对数量优势,随驻留北京时间延长,认知地图结构渐趋复杂;③ 国科大研究生北京城市空间感知表现出偏好特殊性、空间非均衡性、个体差异性、感知过程不完整性等特点;④ 远距离感知、城市空间尺度大、受访者逗留时间短等主客观因素制约了北京城市整体可意象性。     

应用人工神经网络估算伊朗Kharestan流域瞬时顶峰流量(英文)

在水体工程及河流工程项目规划与设计中,瞬时顶峰流量是需要加以了解的最重要因素之一。本研究的目的是,在估算伊朗法尔斯省西北部Kharestan流域的瞬时顶峰流量时,对人工神经网络法与传统方法的功效进行对比。为此,采用了Jamal Beig水文站25年的日顶峰流量和瞬时顶峰流量数据。在Fuller、Sangal及Fill-Steiner经验方法以及人工神经网络方法的基础上进行了估算,并根据RMSE和R2进行了对比。结果显示,采用人工神经网络法的估算值比经验方法的更为精确,RMSE =13.710,R2= 0.942。这表明人工神经网络法比经验方法的误差更低。     

Representing geographical objects with scale‐induced indeterminate boundaries: A neural network‐based data model

The degree of uncertainty of many geographical objects has long been known to be in intimate relation with the scale of its observation and representation. Yet, the explicit consideration of scaling operations when modeling uncertainty is rarely found. In this study, a neural network‐based data model was investigated for representing geographical objects with scale‐induced indeterminate boundaries. Two types of neural units, combined with two types of activation function, comprise the processing core of the model, where the activation function can model either hard or soft transition zones. The construction of complex fuzzy regions, as well as lines and points, is discussed and illustrated with examples. It is shown how the level of detail that is apparent in the boundary at a given scale can be controlled through the degree of smoothness of each activation function. Several issues about the practical implementation of the model are discussed and indications on how to perform complex overlay operations of fuzzy maps provided. The model was illustrated through an example of representing multi‐resolution, sub‐pixel maps that are typically derived from remote sensing techniques.     

URBAN COGNITIVE MAPS: COMPUTATION AND STRUCTURE

Geographic research has not focused on the information we encode into our memories about the characteristics of locations. The accuracy of information encoded in urban cognitive maps, how information varies among individuals, and the structure of the information were studied. Experimental results indicated subjects were more accurate and consistent for socio-economic than for life-cycle characteristics. Some characteristics of the population in a neighborhood, such as race and income, may be more visible and easier to encode through direct contact than other less visible characteristics such as age. A theoretical argument is made that characteristics not encoded for locations can possibly be computed through their covariation with known characteristics.     

Using Web Maps to Analyze the Construction of Global Scale Cognitive Maps

Game-based Web sites and applications are changing the ways in which students learn the world map. In this study, a Web map-based digital learning tool was used as a study aid for a university-level geography course in order to examine the way in which global scale cognitive maps are constructed. A network analysis revealed that clicks were negatively correlated with familiarity, but were positively correlated with area. More significantly, well-known countries did not act as central nodes within network-based mental models, but served a more complex mediating role in structuring the space between subregions.     

Knowledge discovery from soil maps using inductive learning

This paper develops a knowledge discovery procedure for extracting knowledge of soil-landscape models from a soil map. It has broad relevance to knowledge discovery from other natural resource maps. The procedure consists of four major steps: data preparation, data preprocessing, pattern extraction, and knowledge consolidation. In order to recover true expert knowledge from the error-prone soil maps, our study pays specific attention to the reduction of representation noise in soil maps. The data preprocessing step has exhibited an important role in obtaining greater accuracy. A specific method for sampling pixels based on modes of environmental histograms has proven to be effective in terms of reducing noise and constructing representative sample sets. Three inductive learning algorithms, the See5 decision tree algorithm, Naïve Bayes, and artificial neural network, are investigated for a comparison concerning learning accuracy and result comprehensibility. See5 proves to be an accurate method and produces the most comprehensible results, which are consistent with the rules (expert knowledge) used in producing the soil map. The incorporation of spatial information into the knowledge discovery process is found not only to improve the accuracy of the extracted knowledge, but also to add to the explicitness and extensiveness of the extracted soil-landscape model.     

Automatic alignment of contemporary vector data and georeferenced historical maps using reinforcement learning

ABSTRACT

With large amounts of digital map archives becoming available, automatically extracting information from scanned historical maps is needed for many domains that require long-term historical geographic data. Convolutional Neural Networks (CNN) are powerful techniques that can be used for extracting locations of geographic features from scanned maps if sufficient representative training data are available. Existing spatial data can provide the approximate locations of corresponding geographic features in historical maps and thus be useful to annotate training data automatically. However, the feature representations, publication date, production scales, and spatial reference systems of contemporary vector data are typically very different from those of historical maps. Hence, such auxiliary data cannot be directly used for annotation of the precise locations of the features of interest in the scanned historical maps. This research introduces an automatic vector-to-raster alignment algorithm based on reinforcement learning to annotate precise locations of geographic features on scanned maps. This paper models the alignment problem using the reinforcement learning framework, which enables informed, efficient searches for matching features without pre-processing steps, such as extracting specific feature signatures (e.g. road intersections). The experimental results show that our algorithm can be applied to various features (roads, water lines, and railroads) and achieve high accuracy.     

A spatial zoning approach to calibrate and validate urban growth models

Calibration and validation of models predicting urban growth have been largely developed using internal variables. Further investigation is required to improve model’s calibration and validation mixing internal and external variables. To reach this objective, a spatial zoning approach simulating long-term expansion of Mashhad, the second largest city of Iran, was presented in this study. Spatial zoning approaches distinguish local-scale urban dynamics in districts with different socioeconomic characteristics. Thiessen polygons were used to identify districts with different morphology and functional attributes. Urban growth was subsequently simulated for each district using a Multi-Layer Perceptron (MLP) neural network and Markov chains (MC) analysis. MLP and MC algorithms were respectively used to derive transition maps from non-urban to urban use of land and to determine spatial evolution of built-up areas at the metropolitan scale. Results of simulations based on spatial zoning were compared with outcomes of traditional urban growth models. Spatial zoning improved significantly model’s accuracy in respect to more traditional simulation modes. The approach proposed here is appropriate when simulating land-use changes under discontinuous urban expansion.     

Executive Committee Meeting,Washington, D.C. January 28–29, 1972

This study compares how humans and neural networks classify climate types. Human subjects were asked to classify climates from monthly temperature and precipitation patterns. To model their learning process, the same data were used to produce input vectors that trained a pattern associator neural network. Both human subjects and the neural network classified climates accurately after 10 rounds of supervised learning. The neural network successfully modeled the rate of human learning and the ability to learn specific climate categories. Moreover, the neural network weights used to classify climates correspond to distinct visual characteristics in temperature and precipitation. These results suggest that neural networks can model the formation of visual categories.     

Using neural networks and cellular automata for modelling intra‐urban land‐use dynamics

Empirical models designed to simulate and predict urban land‐use change in real situations are generally based on the utilization of statistical techniques to compute the land‐use change probabilities. In contrast to these methods, artificial neural networks arise as an alternative to assess such probabilities by means of non‐parametric approaches. This work introduces a simulation experiment on intra‐urban land‐use change in which a supervised back‐propagation neural network has been employed in the parameterization of several biophysical and infrastructure variables considered in the simulation model. The spatial land‐use transition probabilities estimated thereof feed a cellular automaton (CA) simulation model, based on stochastic transition rules. The model has been tested in a medium‐sized town in the Midwest of São Paulo State, Piracicaba. A series of simulation outputs for the case study town in the period 1985–1999 were generated, and statistical validation tests were then conducted for the best results, based on fuzzy similarity measures.     

Comparing Human and Neural Network Learning of Climate Categories

This study compares how humans and neural networks classify climate types. Human subjects were asked to classify climates from monthly temperature and precipitation patterns. To model their learning process, the same data were used to produce input vectors that trained a pattern associator neural network. Both human subjects and the neural network classified climates accurately after 10 rounds of supervised learning. The neural network successfully modeled the rate of human learning and the ability to learn specific climate categories. Moreover, the neural network weights used to classify climates correspond to distinct visual characteristics in temperature and precipitation. These results suggest that neural networks can model the formation of visual categories.     

多维邻近下浙江城市创新网络演化及其机制研究

以浙江90个县级城市（城区）为对象,基于大数据专利文本挖掘析取城市间合作申请专利数,构建创新网络,采用空间网络模型和负二项回归模型分析2007—2017年浙江城市创新网络的网络结构、时空演化及创新合作强度的影响机制。研究发现：① 浙江城市创新网络规模逐渐扩张,“小世界性”显著,通达性较好,整体呈以杭州湾区为核心的“网络局部化、辐射中心化”特征,等级层次性清晰;② 浙江城市创新合作强度与经济规模、教育水平、政策支持、技术势差、城市行政等级、技术邻近性、边界相邻效应显著正相关且受地理距离约束,认知邻近性和制度邻近性与创新合作间分别呈“U”型和倒“U”型曲线关系,网络效应更多在整体网层面上促进了创新合作。在特定省份县级城市层面探讨了如何加强城市创新网络协同效应,以促进地方城市间的创新联系。     

西安入境游客目的地空间意象认知序列研究

选取西安入境游客为研究对象,依托324 份调查问卷中所提取和统计的认知地图相关数据,分析了西安城市目的地空间意象特征,并重点探讨了认知地图类型及构成要素在游客停留天数上的动态变化过程,据此提出了入境游客目的地城市空间意象认知过程。结果显示：① 西安入境游客的认知地图共分为四大类型,其中以空间型为主,单体型、序列型和混合型次之。② 从空间意象认知要素上看,标志物出现频率最高,其次为区域、边界、节点和道路。以钟楼为中心结合城墙以及东西南北4 条大街,构成了入境游客西安城市目的地空间意象的基本框架。③ 西安入境游客目的地空间认知过程,在认知地图主导类型上,呈现“空间型+单体型→空间型→序列型（混合型）→单体型”的演变序列,在空间认知主导要素上,则为“标志物→标志物+道路→标志物”的发展序列。这一空间认知过程反映了游客对目的地空间意象的关注重点由个性到结构,再到意义的转换。     

Going beyond location: Thematic maps in the early elementary grades

Abstract

Cartographic research has long suggested that preschool and early elementary children possess advanced mapping behaviors in terms of interpreting, making, and using locational and wayfinding maps. To improve instruction in geographic concepts, it is necessary to see if these advanced mapping behaviors apply to thematic or special purpose maps as well, as an understanding of these types of maps is integral to the five fundamental themes of geography. An experiment was conducted to evaluate the thematic mapping abilities of second-grade students. Students viewed maps at the neighborhood, city, and national scales displaying different types of symbolization and presented in a random order. Results indicated that the map tasks were appropriate for this age level, and a statistical test of reaction times did not identify a best symbol type or scale, as all were appropriate for this age level. Responses to discussion questions indicated the students also have the ability to describe spatial patterns. The interpretation of quantitative information on maps is within the ability level of second grade students, and materials currently used in second grade should be revised to include these maps.     

Putting together the pieces of the puzzle: Traces of New Zealand in international school geography textbooks

How does a fairly remote space (the islands of New Zealand) contribute to the development of geographical literacy in the geography textbooks of post‐socialist (Romania), Western (Germany) and post‐colonial (Mexico) countries? A four‐step analysis of three textbooks of international selection revealed different regionalisation cultures combined with traditional and recently implemented educational paradigms of geography education and led to a similar result. Despite the many ways in which New Zealand's geographical distinctiveness could be used to enhance geographical literacy, there are few operationalisations of educational targets that might contribute to sustainable long‐term learning. Further, these results also offer a possible explanation for the absence of the ‘southern lands’ on students' cognitive maps of the world.