The protection and enhancement of visual resources constitute an on-going challenge to the planning authorities in many communities. The crux of this challenge is to guide development towards built and natural landscape forms that will not cause detriment to an existing landscape character. To understand and cope with this problem, there is the need for a means to define and model a landscape's character, to identify methods for constructing that character definition, to create tools for storing and using such a definition to visualize its spatial manifestations, and to incorporate alternative development regulatory parameters in order to assess their impact on landscape character. Current spatial data technologies are able to portray inventories of specific, real-world objects. While well established in the planning profession, these technologies and their attendant data manipulation tools do not easily facilitate the creation of generalized, non-specific statements that are applicable across a region. Such generalized statements regarding visual and spatial features are at the heart of descriptions of landscape character and implicit within most planning regulations intended to produce a desirable landscape character. Current spatial data tools therefore do not satisfy the stated needs of planning for landscape character. In satisfying these conceptual, methodological and technological deficiencies, the research presented in this dissertation defines and demonstrates a theory of landscape grammar which formally draws parallels between the structures of linguistics and the character of landscapes. A landscape grammar defines a landscape character using a spatial vocabulary and syntax rules and can be applied to a site to generate landscape forms that embody the defined character. In this dissertation, the spatial counterparts of the linguistic concepts of vocabulary and grammar rules are formalized and implemented for use in a custom-developed geographic information system. Methods that enable the use of landscape grammars in a planning environment are presented and subsequently applied through the formal expression of planning regulations into the grammar-based model. The theory, methods and software implementation are demonstrated using a residential area of the island of Bermuda. The iterative grammatical generation of an example two-dimensional landscape scene is demonstrated with further three-dimensional representations of the results for visualization purposes. Alternative planning regulations are also incorporated into the case study grammar and resultant three-dimensional landscapes are shown. Several suggestions for future research on landscape grammars are offered in the conclusions of the dissertation.

The protection and enhancement of visual resources constitute an on-going challenge to the planning authorities in many communities. The crux of this challenge is to guide development towards built and natural landscape forms that will not cause detriment to an existing landscape character. To understand and cope with this problem, there is the need for a means to define and model a landscape's character, to identify methods for constructing that character definition, to create tools for storing and using such a definition to visualize its spatial manifestations, and to incorporate alternative development regulatory parameters in order to assess their impact on landscape character. Current spatial data technologies are able to portray inventories of specific, real-world objects. While well established in the planning profession, these technologies and their attendant data manipulation tools do not easily facilitate the creation of generalized, non-specific statements that are applicable across a region. Such generalized statements regarding visual and spatial features are at the heart of descriptions of landscape character and implicit within most planning regulations intended to produce a desirable landscape character. Current spatial data tools therefore do not satisfy the stated needs of planning for landscape character. In satisfying these conceptual, methodological and technological deficiencies, the research presented in this dissertation defines and demonstrates a theory of landscape grammar which formally draws parallels between the structures of linguistics and the character of landscapes. A landscape grammar defines a landscape character using a spatial vocabulary and syntax rules and can be applied to a site to generate landscape forms that embody the defined character. In this dissertation, the spatial counterparts of the linguistic concepts of vocabulary and grammar rules are formalized and implemented for use in a custom-developed geographic information system. Methods that enable the use of landscape grammars in a planning environment are presented and subsequently applied through the formal expression of planning regulations into the grammar-based model. The theory, methods and software implementation are demonstrated using a residential area of the island of Bermuda. The iterative grammatical generation of an example two-dimensional landscape scene is demonstrated with further three-dimensional representations of the results for visualization purposes. Alternative planning regulations are also incorporated into the case study grammar and resultant three-dimensional landscapes are shown. Several suggestions for future research on landscape grammars are offered in the conclusions of the dissertation.

While machine learning (ML) methods have achieved considerable success on several mainstream problems in vision and language modeling, they are still challenged by their lack of interpretable decision-making that is consistent with scientific knowledge, limiting their applicability for scientific discovery applications. Recently, a new field of machine learning that infuses domain knowledge into data-driven ML approaches, termed Knowledge-Guided Machine Learning (KGML), has gained traction to address the challenges of traditional ML. Nonetheless, the inner workings of KGML models and algorithms are still not fully understood, and a better comprehension of its advantages and pitfalls over a suite of scientific applications is yet to be realized. In this thesis, I first tackle the task of understanding the role KGML plays at shaping the outputs of a neural network, including its latent space, and how such influence could be harnessed to achieve desirable properties, including robustness, generalizability beyond training data, and capturing knowledge priors that are of importance to experts. Second, I use and further develop loss landscape visualization tools to better understand ML model optimization at the network parameter level. Such an understanding has proven to be effective at evaluating and diagnosing different model architectures and loss functions in the field of KGML, with potential applications to a broad class of ML problems. / Doctor of Philosophy / My research aims to address some of the major shortcomings of machine learning, namely its opaque decision-making process and the inadequate understanding of its inner workings when applied in scientific problems. In this thesis, I address some of these shortcomings by investigating the effect of supplementing the traditionally data-centric method with human knowledge. This includes developing visualization tools that make understanding such practice and further advancing it easier. Conducting this research is critical to achieving wider adoption of machine learning in scientific fields as it builds up the community's confidence not only in the accuracy of the framework's results, but also in its ability to provide satisfactory rationale.

Effective coastal planning is inclusive and incorporates the variety of user needs, values, and interests associated with coastal environments. This requires understanding how people relate to coastal environments as ‘places’, imbued with values and meanings, and accordingly, tools that can capture place and connect with people’s ‘sense of place’ have the potential for supporting effective coastal management strategies. Realistic, immersive geographical visualizations, i.e., geovisualizations, theoretically hold potential to serve such a role in coastal planning. However, significant research gaps exist around this application context. Firstly, place theory and geovisualizations are rarely explicitly linked in the same studies, leaving questions around the (potential) relationship between these tools and sense of place. Secondly, geovisualization work has focused on terrestrial environments, and research on how to realistically model coastal places is currently in its infancy. This dissertation aims to address these gaps by pursuing two research objectives. The first objective is to explore the ‘human component’ of geovisualizations, referring to how these tools operate within the social and cultural dimensions germane to environmental management plans and processes. In accordance with the discussion above, this exploration is framed through place theories and concepts, and regards realistic geovisualizations as ‘place-based’ tools. The second objective concerns the coastal context, and it involves elucidating the considerations around developing and using terrestrial-to-marine geovisualizations as tools for inclusive coastal planning and management. The dissertation is composed of five manuscripts, which have been prepared as standalone articles for submission to academic journals. Each manuscript details a study designed to support an aspect of the research objectives, respectively serving (1) to develop a theory of geovisualizations as place-based tools, (2) to explore the theory in the coastal context, (3) to examine the relationship between sense of place and one’s mental visualization of place, (4) to develop a coastal geovisualization under place-based considerations and examine its capacity for connecting to sense of place, and (5) to assess the geovisualization’s potential as a tool for inclusive coastal planning efforts. The first and second study consist of literature review work. The third study involves a survey administered to residents of the Capital Regional District, which collected data for examining a potential relationship between the way people visualize coastal places and how they value and relate to these places. The fourth and fifth study involve developing a coastal geovisualization of Sidney Spit, and then employing focus groups to examine its ability for connecting with people’s sense of place (i.e., fourth study) and utility as a tool for inclusive planning (i.e., fifth study). Outcomes from the first study include a theory on how geovisualizations can function as place-based tools, and this was developed by integrating place concepts with ideas and conceptual models from human-media interaction and sense of presence research. The second study produced insight on how values and interests of different coastal user groups can influence understandings and perceptions of coastal places, and it used this insight to develop recommendations for coastal geovisualizations - full navigability, dynamic elements, and flexibility (i.e., allowing for continual modification and scenario building). The third study produced empirical evidence that place-based values and interests (i.e., framed through sense of place and concerns for place) can influence one’s mental visualization of place in terms of the types of elements people include and perspectives they take in said visualization. The fourth study demonstrated that the presence of certain elements in coastal geovisualizations (such as people, dogs, birds, marine life, vegetation, and boats) can contribute to realism and sense of place; however, simultaneously, deficiencies in numbers and varieties of these elements can detract from realism and sense of place. In addition, the fourth study found that the incorporation of soundscape and viewshed elements is significant for the tool’s ability to connect with sense of place. The fifth study demonstrated the geovisualization’s usefulness for assessing certain qualities of management scenarios, such as aesthetics and functionality of fencing around a restoration area and potential viewshed impacts associated with locations of moored boats. The study also found that incorporating navigability into the geovisualization proved to be valuable for enhancing understandings around scenarios that hold implications for the marine environment because it allowed users to cross the land-sea interface and experience underwater places. / Graduate

No description available.

Parmi tous les jeux de société joués par les humains au cours de l’histoire, le jeu de go était considéré comme l’un des plus difficiles à maîtriser par un programme informatique [Van Den Herik et al., 2002]; Jusqu’à ce que ce ne soit plus le cas [Silveret al., 2016]. Cette percée révolutionnaire [Müller, 2002, Van Den Herik et al., 2002] fût le fruit d’une combinaison sophistiquée de Recherche arborescente Monte-Carlo et de techniques d’apprentissage automatique pour évaluer les positions du jeu, mettant en lumière le grand potentiel de l’apprentissage automatique pour résoudre des jeux. L’apprentissage antagoniste, un cas particulier de l’optimisation multiobjective, est un outil de plus en plus utile dans l’apprentissage automatique. Par exemple, les jeux à deux joueurs et à somme nulle sont importants dans le domain des réseaux génératifs antagonistes [Goodfellow et al., 2014] ainsi que pour maîtriser des jeux comme le Go ou le Poker en s’entraînant contre lui-même [Silver et al., 2017, Brown andSandholm, 2017]. Un résultat classique de la théorie des jeux indique que les jeux convexes-concaves ont toujours un équilibre [Neumann, 1928]. Étonnamment, les praticiens en apprentissage automatique entrainent avec succès une seule paire de réseaux de neurones dont l’objectif est un problème de minimax non-convexe et non-concave alors que pour une telle fonction de gain, l’existence d’un équilibre de Nash n’est pas garantie en général. Ce travail est une tentative d'établir une solide base théorique pour l’apprentissage dans les jeux. La première contribution explore le théorème minimax pour une classe particulière de jeux non-convexes et non-concaves qui englobe les réseaux génératifs antagonistes. Cette classe correspond à un ensemble de jeux à deux joueurs et a somme nulle joués avec des réseaux de neurones. Les deuxième et troisième contributions étudient l’optimisation des problèmes minimax, et plus généralement, les inégalités variationnelles dans le cadre de l’apprentissage automatique. Bien que la méthode standard de descente de gradient ne parvienne pas à converger vers l’équilibre de Nash de jeux convexes-concaves simples, il existe des moyens d’utiliser des gradients pour obtenir des méthodes qui convergent. Nous étudierons plusieurs techniques telles que l’extrapolation, la moyenne et la quantité de mouvement à paramètre négatif. La quatrième contribution fournit une étude empirique du comportement pratique des réseaux génératifs antagonistes. Dans les deuxième et troisième contributions, nous diagnostiquons que la méthode du gradient échoue lorsque le champ de vecteur du jeu est fortement rotatif. Cependant, une telle situation peut décrire un pire des cas qui ne se produit pas dans la pratique. Nous fournissons de nouveaux outils de visualisation afin d’évaluer si nous pouvons détecter des rotations dans comportement pratique des réseaux génératifs antagonistes. / Among all the historical board games played by humans, the game of go was considered one of the most difficult to master by a computer program [Van Den Heriket al., 2002]; Until it was not [Silver et al., 2016]. This odds-breaking break-through [Müller, 2002, Van Den Herik et al., 2002] came from a sophisticated combination of Monte Carlo tree search and machine learning techniques to evaluate positions, shedding light upon the high potential of machine learning to solve games. Adversarial training, a special case of multiobjective optimization, is an increasingly useful tool in machine learning. For example, two-player zero-sum games are important for generative modeling (GANs) [Goodfellow et al., 2014] and mastering games like Go or Poker via self-play [Silver et al., 2017, Brown and Sandholm,2017]. A classic result in Game Theory states that convex-concave games always have an equilibrium [Neumann, 1928]. Surprisingly, machine learning practitioners successfully train a single pair of neural networks whose objective is a nonconvex-nonconcave minimax problem while for such a payoff function, the existence of a Nash equilibrium is not guaranteed in general. This work is an attempt to put learning in games on a firm theoretical foundation. The first contribution explores minimax theorems for a particular class of nonconvex-nonconcave games that encompasses generative adversarial networks. The proposed result is an approximate minimax theorem for two-player zero-sum games played with neural networks, including WGAN, StarCrat II, and Blotto game. Our findings rely on the fact that despite being nonconcave-nonconvex with respect to the neural networks parameters, the payoff of these games are concave-convex with respect to the actual functions (or distributions) parametrized by these neural networks. The second and third contributions study the optimization of minimax problems, and more generally, variational inequalities in the context of machine learning. While the standard gradient descent-ascent method fails to converge to the Nash equilibrium of simple convex-concave games, there exist ways to use gradients to obtain methods that converge. We investigate several techniques such as extrapolation, averaging and negative momentum. We explore these techniques experimentally by proposing a state-of-the-art (at the time of publication) optimizer for GANs called ExtraAdam. We also prove new convergence results for Extrapolation from the past, originally proposed by Popov [1980], as well as for gradient method with negative momentum. The fourth contribution provides an empirical study of the practical landscape of GANs. In the second and third contributions, we diagnose that the gradient method breaks when the game’s vector field is highly rotational. However, such a situation may describe a worst-case that does not occur in practice. We provide new visualization tools in order to exhibit rotations in practical GAN landscapes. In this contribution, we show empirically that the training of GANs exhibits significant rotations around Local Stable Stationary Points (LSSP), and we provide empirical evidence that GAN training converges to a stable stationary point, which is a saddle point for the generator loss, not a minimum, while still achieving excellent performance.