Downloads & Free Reading Options - Results

Zoomable full resolution image available at davidrumsey.com . This image 14359.191 is a part of image group 14359.000 . Twelve statistical maps and chart representing the United States, featuring the geographic density of nurses, hospital beds and patients in hospitals in 1965, as well as the percentage of health expenses per person in 1962, and the percentage of the population consulting physicians and dentists, and missing work or school in 1964 and 1965. Maps also show political boundaries, bodies of water, coastlines and islands. Maps include legends and explanatory notes, as well as bar scales and scale statements (scales differ). With two inset maps within each map: Principal islands of Hawaii -- Alaska. Chart shows hospital statistics in 1965. Colored lithograph. Together, chart and maps are 42 x 31 cm, on sheet 49 x 35 cm. Chart and maps appear in Special subject maps section, subsection Socio-Cultural. The national atlas of the United States of America, by the United States Geological Survey; published in Washington D. C., 1970. Bound in navy blue board, with title printed in silver on both front cover and spine. Accompanied by envelope with six overlay sheets, tucked between final page and back cover of volume. Collation: [i-vi], vii-xiii, [1], 2-417, A1, A2, B1, B2, C, D. Atlas contains 770 maps and 18 charts. Includes a dedication, foreward, list of contributors, introduction, table of contents and index to map subjects. Topic covered: physical geography, history, economics, culture, administrative boundaries and cartography. In addition, atlas also provides maps of the world, as related to the United States. Maps show political boundaries, cities, railways, roads, topography, bodies of water, glaciers, drainage, coastlines, islands, water depths and time zones. Topical maps feature other details, such as history, geology, climate, agriculture, population, racial demographics (including indigenous peoples), religion, language and transportation. Some maps use data visualization to further illustrate geographical information, with charts overlaid upon the landscape. "Adapted from "About The National Atlas of the United States of America," by the U.S. Geological Survey: The National Atlas of the United States of America was published in 1970. It was designed to be of practical use to decision makers in government and business, and for planners and research scholars as well as others needing to visualize country-wide distributional patterns and relationships between environmental phenomena and human activities. The National Atlas represents the principal characteristics of the country in 1970, including its physical features, historical evolution, economic activities, sociocultural conditions, administrative subdivisions, and place in world affairs. Various federal agencies, professional organizations, and commercial firms had advocated producing a National Atlas of the United States of America, but the magnitude of the task and the scope of the research required deterred those who would begin it. Late in 1954, the National Academy of Sciences-National Research Council established a Committee on the National Atlas of the United States, with representatives from several federal mapmaking agencies. The committee's primary responsibilities were to coordinate all federal agencies that would be involved in producing the atlas and to ensure uniform quality in its cartography. This proved to be a nearly impossible task, and consequently in 1961 the committee terminated itself. In so doing it recommended that the atlas be completed by one federal agency, preferably the Geological Survey in the U.S. Department of the Interior. In March of 1961, the Secretary of the Interior accepted the challenge. Congress appropriated funds to begin work on the National Atlas in 1963, and on reimbursable loan the Library of Congress made available the chief of its Geography and Map Division, Dr. Arch C. Gerlach, to serve as editor. Eighty-four agencies and bureaus appointed liaison officers to the National Atlas Project, base maps were prepared at four scales, and an advisory group of eminent cartographers and geographers collaborated to formulate fundamental design principles and specifications. The first part of the National Atlas is devoted to general reference maps that contain most of the forty-one thousand place names recorded in the index. These maps were included for the convenience of readers wanting basic locational information. In the thematic section of the National Atlas, separate subdivisions deal with the country's physical, historical, economic, and socio-cultural characteristics. The maps in this portion of the atlas represent the relationships between human beings and their environment while offering scientific bases for analyzing the nation's economic development in 1970. The National Atlas proved to be the last paper atlas of this magnitude produced by the federal government."

We consider the problem of offline, pool-based active semi-supervised learning on graphs. This problem is important when the labeled data is scarce and expensive whereas unlabeled data is easily available. The data points are represented by the vertices of an undirected graph with the similarity between them captured by the edge weights. Given a target number of nodes to label, the goal is to choose those nodes that are most informative and then predict the unknown labels. We propose a novel framework for this problem based on our recent results on sampling theory for graph signals. A graph signal is a real-valued function defined on each node of the graph. A notion of frequency for such signals can be defined using the spectrum of the graph Laplacian matrix. The sampling theory for graph signals aims to extend the traditional Nyquist-Shannon sampling theory by allowing us to identify the class of graph signals that can be reconstructed from their values on a subset of vertices. This approach allows us to define a criterion for active learning based on sampling set selection which aims at maximizing the frequency of the signals that can be reconstructed from their samples on the set. Experiments show the effectiveness of our method.

 ModelStream with Bert de Vries, Thijs van de Laar and Dmitry Bagaev ( https://biaslab.github.io/ ): "Implementing Active Inference by Message Passing in a Factor Graph" Active Inference Institute information: Website: https://activeinference.org/ Twitter: https://twitter.com/InferenceActive Discord: https://discord.gg/8VNKNp4jtx YouTube: / activeinference Active Inference Livestreams: https://coda.io/@active-inference-ins... CSID: d1fbd7c4a6cb758d Content Managed by ContentSafe.co

This paper investigates the problem of active learning for binary label prediction on a graph. We introduce a simple and label-efficient algorithm called S2 for this task. At each step, S2 selects the vertex to be labeled based on the structure of the graph and all previously gathered labels. Specifically, S2 queries for the label of the vertex that bisects the *shortest shortest* path between any pair of oppositely labeled vertices. We present a theoretical estimate of the number of queries S2 needs in terms of a novel parametrization of the complexity of binary functions on graphs. We also present experimental results demonstrating the performance of S2 on both real and synthetic data. While other graph-based active learning algorithms have shown promise in practice, our algorithm is the first with both good performance and theoretical guarantees. Finally, we demonstrate the implications of the S2 algorithm to the theory of nonparametric active learning. In particular, we show that S2 achieves near minimax optimal excess risk for an important class of nonparametric classification problems.

Successful navigation in day-to-day activities fundamentally requires the encoding of the spatial layout of one’s surroundings. When arriving at a new location, making decisions about how or where to explore could play a key role in aiding the translation of one’s physical surroundings into an internal spatial representation that can be modified and refined with further navigation experience. We previously found an advantage of active decision-making (vs. being guided through a new environment) in the acquisition of spatial graph knowledge (i.e., a flexible understanding of specific locations within an environment and the path connections between them; Chrastil & Warren, 2015), but not for map-like survey knowledge (Chrastil & Warren, 2013). However, more research is warranted to understand the specific cognitive advantages behind active decision-making in the acquisition of spatial graph knowledge. We theorize that making decisions allows people to make small predictions about what they think they will encounter during exploration, and then get feedback on those decisions. This reinforcement learning-inspired hypothesis suggests that those who are guided could benefit from making explicit predictions. Alternatively, it is possible that simply the act of choosing ones path is the strongest factor in active decision-making. The following study aims to systematically investigate the effects of decision-making on spatial graph learning during spatial exploration by including the factor of making explicit predictions about the locations of salient features in an environment, followed by feedback, and the subsequent reinforcement learning that ensues from this. We examine the interplay between decision-making and reinforcement learning using a between-subjects, 2 (Decision-Making: Active Exploration, Passive Exploration) x 2 (Reinforcement Learning: Making Explicit Predictions and Feedback, Making Selections) Factorial Design throughout an exploration phase, followed by a testing (i.e., wayfinding trials) phase using immersive walking virtual reality with an HTC VIVE Pro Eye VR head-mounted display. First, participants are randomly assigned to one of four exploration phase groups: 1) active exploration, making predictions followed by feedback; 2) active exploration, making selections; 3) passive exploration, making predictions, followed by feedback; 4) passive exploration, making selections. Participants in the explicit prediction and feedback groups will first enter an alcove and only see a 3x3 selection grid (i.e., without seeing the hidden target object that is placed in this alcove). Each cell in the 3x3 grid displays an image and name tag of one of 9 possible target objects located in the maze. Using a handheld controller, these participants will click the controller on one of the 9 cells to predict what object they think is in the alcove. After making their prediction, the selection grid will disappear, and the hidden target object placed in this alcove will be revealed as feedback. Initially, participants may make unsystematic predictions of the target objects’ locations as they explore alcoves for the first time, as there are no explicit cues that would suggest one target object belonging to a certain alcove. Through further exploration of the virtual maze, these participants may begin to make more systematic predictions, which when followed by feedback may improve their learning of the 9 target objects’ locations. Conversely, participants in the selection group will see both a 3x3 selection grid and a visible target object when arriving at alcoves. Rather than making an explicit prediction and receiving feedback, these participants simply select the appropriate cell that matches the presented target object. With regards to decision-making factor, participants in one of the two active exploration conditions can freely explore without any bias or predetermined path, while participants in one of the two passive exploration conditions are yoked (i.e., an experimenter will guide passive participants using a handheld controller on the same exploration trajectory of an active explorer with their respective reinforcement learning condition, allowing them to experience the same visual stimuli). Participants in both passive groups are still instructed to use the 3x3 selection grid to either make predictions and receive feedback or make selections. Participants in all four groups are tasked with locating and learning the locations of 9 target objects for approximately seven minutes. Subsequently, all participants across the exploration groups are tested on the acquisition of their spatial graph knowledge in the test phase. On each trial, participants are instructed to travel from a starting object to a target object using the maze’s hallways, however, all of the objects are turned into red spheres to minimize feedback, so they do not necessarily know if they reached the correct target. There are 36 test trials, with a time limit for each trial. The primary dependent variable of interest is wayfinding success, which is the proportion of correct trials that participants correctly reach the specified target. Overall, reinforcement learning during the encoding of novel spatial environments may facilitate aspects of decision-making and hence represent one of the most critical components of active spatial learning that may contribute to better spatial knowledge acquisition. This knowledge can, in turn, serve as a method for training individuals to improve their navigation ability.

In graph-based active learning, algorithms based on expected error minimization (EEM) have been popular and yield good empirical performance. The exact computation of EEM optimally balances exploration and exploitation. In practice, however, EEM-based algorithms employ various approximations due to the computational hardness of exact EEM. This can result in a lack of either exploration or exploitation, which can negatively impact the effectiveness of active learning. We propose a new algorithm TSA (Two-Step Approximation) that balances between exploration and exploitation efficiently while enjoying the same computational complexity as existing approximations. Finally, we empirically show the value of balancing between exploration and exploitation in both toy and real-world datasets where our method outperforms several state-of-the-art methods.

http://uf.catalog.fcla.edu/uf.jsp?st=UF021521928&ix=pm&I=0&V=D&pm=1

To train good supervised and semi-supervised object classifiers, it is critical that we not waste the time of the human experts who are providing the training labels. Existing active learning strategies can have uneven performance, being efficient on some datasets but wasteful on others, or inconsistent just between runs on the same dataset. We propose perplexity based graph construction and a new hierarchical subquery evaluation algorithm to combat this variability, and to release the potential of Expected Error Reduction. Under some specific circumstances, Expected Error Reduction has been one of the strongest-performing informativeness criteria for active learning. Until now, it has also been prohibitively costly to compute for sizeable datasets. We demonstrate our highly practical algorithm, comparing it to other active learning measures on classification datasets that vary in sparsity, dimensionality, and size. Our algorithm is consistent over multiple runs and achieves high accuracy, while querying the human expert for labels at a frequency that matches their desired time budget.

This article is from Frontiers in Neuroinformatics , volume 8 . Abstract The aim in high-resolution connectomics is to reconstruct complete neuronal connectivity in a tissue. Currently, the only technology capable of resolving the smallest neuronal processes is electron microscopy (EM). Thus, a common approach to network reconstruction is to perform (error-prone) automatic segmentation of EM images, followed by manual proofreading by experts to fix errors. We have developed an algorithm and software library to not only improve the accuracy of the initial automatic segmentation, but also point out the image coordinates where it is likely to have made errors. Our software, called gala (graph-based active learning of agglomeration), improves the state of the art in agglomerative image segmentation. It is implemented in Python and makes extensive use of the scientific Python stack (numpy, scipy, networkx, scikit-learn, scikit-image, and others). We present here the software architecture of the gala library, and discuss several designs that we consider would be generally useful for other segmentation packages. We also discuss the current limitations of the gala library and how we intend to address them.