Computing and Information Systems - Theses

Permanent URI for this collection

Search Results

Now showing 1 - 10 of 46
  • Item
    Thumbnail Image
    Practical declarative debugging of mercury programs
    MacLarty, Ian Douglas. (University of Melbourne, 2006)
  • Item
    Thumbnail Image
    Practical declarative debugging of mercury programs
    MacLarty, Ian Douglas. (University of Melbourne, 2006)
  • Item
    Thumbnail Image
    Generating Deep Network Explanations with Robust Attribution Alignment
    Zeng, Guohang ( 2021)
    Deep Neural Networks (DNNs) have achieved impressive success in many fields, yet the black-box nature of DNNs hinders their application in critical domains, such as the medical domain. To this end, Interpretable Machine Learning (IML) is a research field aims to understand the mechanism behind DNNs via interpretability methods, which aim to provide explanations to human users and help them understand how black-box models make decisions. Current IML methods produce post-hoc attribution maps on pre-trained models. However, recent studies have shown that most of these methods yield unfaithful and noisy explanations. In this study, we present a new paradigm of interpretability methods to improve the quality of explanations. We treat a model’s explanations as a part of the network’s outputs, then generate attribution maps from the underlying deep network. The generated attribution maps are up-sampled from the last convolutional layer of the network to obtain localization information about the target to be explained. Another intuition behind this study is leveraging the connection between interpretability and adversarial machine learning to improve the quality of explanations. Inspired by recent studies that showed adversarially robust models’ saliency aligns well with human perception, we utilize attribution maps from the robust model to supervise the learned attributions. Our proposed method can produce visually plausible explanations along with the prediction in inference phase. Experiments on real datasets show that our proposed method yields more faithful explanations than post-hoc attribution methods with lighter computational costs.
  • Item
    Thumbnail Image
    Foundations for reasoning about holistic specifications
    Nguyen, Duc Than ( 2020)
    Specifications of sufficient conditions may be enough for reasoning about complete and unchanging programs of a closed system. Nevertheless, there is no luxury of trusting external components of probably unknown provenance in an open world that may be buggy or potentially malicious. It is critical to ensure that our components are robust when cooperating with a wide variety of external components. Holistic specifications, which are concerned with sufficient and necessary conditions, could make programs more robust in an open-world setting. In this thesis, we lay the foundations for reasoning about holistic specifications. We give an Isabelle/HOL mechanization of holistic specifications focusing on object-based programs. We also pave a way to reason about holistic specifications via proving some key lemmas that we hope will be useful in the future to establish a general logic for holistic specifications.
  • Item
    Thumbnail Image
    Learning to generalise through features
    Grebenyuk, Dmitry ( 2020)
    A Markov decision process (MDP) cannot be used for learning end-to-end control policies in Reinforcement Learning when the dimension of the feature vectors changes from one trial to the next. For example, this difference is present in an environment where the number of blocks to manipulate can vary. Because we cannot learn a different policy for each number of blocks, we suggest framing the problem as a POMDP instead of the MDP. It allows us to construct a constant observation space for a dynamic state space. There are two ways we can achieve such construction. First, we can design a hand-crafted set of observations for a particular problem. However, that set cannot be readily transferred to another problem, and it often requires domain-dependent knowledge. On the other hand, a set of observations can be deduced from visual observations. This approach is universal, and it allows us to easily incorporate the geometry of the problem into the observations, which can be challenging to hard-code in the former method. In this Thesis, we examine both of these methods. Our goal is to learn policies that can be generalised to new tasks. First, we show that a more general observation space can improve the performance of policies tested in untrained tasks. Second, we show that meaningful feature vectors can be obtained from visual observations. If properly regularised, these vectors can reflect the spacial structure of the state space and used for planning. Using these vectors, we construct an auto-generated reward function, able to learn working policies.
  • Item
    Thumbnail Image
    Embedding Graphs for Shortest-Path Distance Predictions
    Zhao, Zhuowei ( 2020)
    Graph is an important data structure and is used in an abundance of real-world applications including navigation systems, social networks, and web search engines, just to name but a few. We study a classic graph problem – computing graph shortest-path distances. This problem has many applications, such as finding nearest neighbors for place of interest(POI) recommendation or social network friendship recommendation. To compute a shortest-path distance, traditional approaches traverse the graph to find the shortest path and return the path length. These approaches lack time efficiency over large graphs. In the applications above, the distances may be needed first (e.g., to rank POIs), while the actual shortest paths may be computed later (e.g., after a POI has been chosen). Thus, an alternative approach precomputes and stores the distances, and answers distance queries with simple lookups. This approach, however, falls short in the space cost – O(n^2) in the worst-case for vertices, even with various optimizations. To address these limitations, we take an embedding based approach to predict the shortest-path distance between two vertices using their embeddings without computing their path online or storing their distance offline. Graph embedding is an emerging technique for graph analysis that has yielded strong performance in applications such as node classification, link prediction, graph reconstruction, and more. We propose a representation learning approach to learn a k-dimensional (k<
  • Item
    Thumbnail Image
    What you get is what you see: Decomposing Epistemic Planning using Functional STRIPS
    Hu, Guang ( 2019)
    Epistemic planning --- planning with knowledge and belief --- is essential in many multi-agent and human-agent interaction domains. Most state-of-the-art epistemic planners solve this problem by compiling to propositional classical planning, for example, generating all possible knowledge atoms, or compiling epistemic formula to normal forms.It is noted that the compilations are typically exponentially larger than the original problem. However, these methods become computationally infeasible as problems grow. In addition, those methods only works on propositional variables in discrete domains. In this thesis, we decompose epistemic planning by delegating epistemic logic reasoning to an external solver. We do this by modelling the problem using \emph{functional STRIPS}, which is more expressive than standard STRIPS and supports the use of external, black-box functions within action models. Exploiting recent work that demonstrates the relationship between what an agent `sees' and what it knows, we allow modellers to provide new implementations of externals functions. These define what agents see in their environment, allowing new epistemic logics to be defined without changing the planner. As a result, the capability and flexibility of the epistemic model itself are increased, as our model is able to avoid exponential pre-compilation steps and handle logics from continuous domains.We ran evaluations on well-known epistemic planning benchmarks to compare with an existing state-of-the-art planner, and on new scenarios based on different external functions. The results show that our planner scales significantly better than the state-of-the-art planner which we compared against, and can express problems more succinctly.
  • Item
    Thumbnail Image
    Towards improving the network architecture of GANs and their evaluation methods
    Barua, Sukarna ( 2019)
    Generative Adversarial Networks (GANs) are a powerful class of generative models. GAN models have recently brought significant success in image synthesis tasks. One key issue concerning GANs is the design of a network architecture that results in high training stability and sample quality. GAN models consist of two distinct neural networks known as the generator and discriminator. Conventional practice is to use a deep convolution architecture for both networks that eliminates fully connected layers from the architecture or restricts their uses to only input and output layers. Our investigation reveals that eliminating fully connected layers from the network architecture of GANs is not the best practice, and more effective GAN architecture can be designed by rather exploiting fully connected layers in the conventional convolution architecture. In this respect, we propose an improved network architecture for GANs that employs multiple fully connected layers in both the generator and discriminator networks. Models based on our proposed architecture learn both faster than the conventional architecture and also generate higher quality of samples. In addition, our proposed architecture demonstrates higher training stability than the conventional architecture in several experimental settings. We demonstrate the effectiveness of our architecture in generating high-fidelity images on four benchmark image datasets. Another key challenge when using GANs is how to best measure their ability to generate realistic data. In this regard, we demonstrate that an intrinsic dimensional characterization of the data space learned by a GAN model leads to an effective evaluation metric for GAN quality. In particular, we propose a new evaluation measure, CrossLID, that assesses the local intrinsic dimensionality (LID) of real-world data with respect to neighborhoods found in GAN-generated samples. Intuitively, CrossLID measures the degree to which manifolds of two data distributions coincide with each other. We compare our proposed measure to several state-of-the-art evaluation metrics. Our experiments show that CrossLID is strongly correlated with the progress of GAN training, is sensitive to mode collapse, is robust to small-scale noise and image transformations, and robust to sample size. One key advantage of the proposed CrossLID metric is the ability to assess mode-wise performance of GAN models. The mode-wise evaluation can be used to assess how well a GAN model has learned the different modes present in the target data distribution. We demonstrate how the proposed mode-wise assessment can be utilized during the GAN training process to detect unlearned modes. This leads us to an effective training strategy for GANs that dynamically mitigate unlearned modes by oversampling them during the training. Experiments on benchmark image datasets show that our proposed training approach achieves better performance scores than the conventional GAN training. In addition, our training approach demonstrates higher stability against mode failures of GANs compared to the conventional training.
  • Item
    Thumbnail Image
    Performances and publics while watching and live-streaming video games on Twitch.tv
    Robinson, Naomi Eleanor Isobel ( 2019)
    Twitch.tv is a video live-streaming website that launched in 2011 with content centred mostly, but not exclusively, on the playing of video games. Streamers or broadcasters play games in real-time often accompanied by a face camera and audio, while viewers or audiences watch them and interact through a text chat. This study responds to the small, but growing literature surrounding Twitch, and addresses the relative lack of ethnographic research on the topic. Previous research on the platform has focussed thus far on technical aspects of the platform, however user-focused qualitative research on the platform has started to emerge, making this research both timely and relevant. This thesis considers how, and to what extent, the social practices of users contribute to the concepts of ‘networked publics’ and ‘social performance’. It draws on the work of danah boyd and Erving Goffman and considers the usefulness of their theoretical contributions to help contextualise the forms and amendments associated with platforms like Twitch. The analysis emerges from an ethnographic study conducted completely online that features reflexive participant observation, semi-structured, open-ended interviews conducted via email, and in-depth observations of participants’ channels. The thesis is divided into three thematically-organised main data chapters that then feed into a discussion that draws them together to consider a larger conceptual framework. The first such data chapter, ‘Twitch as a Social Media Platform’, argues that the platform demonstrates its role as a social networking site through evidence of matchmaking and mental health. The second main chapter, ‘Twitch as a hobby-profession’, addresses casual and serious leisure and considers the platform in terms of personal investment, branding, and streamer motivation. The third main chapter, ‘Interactions of Streamers and Viewers’, considers the different types of interactions displayed between various users including parasocal relationships and how audiences may hold power on Twitch. Overall, the thesis offers insight into platform use and it characterises Twitch as a user-led participatory space for like-minded individuals who interact in particular ways in a shared community of practice. The interactions exist along a flexible continuum of differing levels of intimacy where users can lurk, actively participate, and network on both personal and professional levels. Audiences are critical for the platform to function, for communities to flourish, and for streamer success. Streamers build rapport and construct ‘authentic’ brands to attract viewers and promote loyalty and sincerity, and users are seen to actively shape and shift extant social structures and practices over time. Ultimately, users find meaning, produce a sense of community belonging, forge social networks, and shape their own identities in relation to others. The thesis concludes that Twitch somewhat paradoxically is both fleeting and robustly sustained by its contemporary community of practice. This community is produced and maintained through interaction and performance that shapes the construction of Twitch’s publics, with Twitch itself acting as a large participatory public as well. Performative sociality and networking are understood as key driving forces for Twitch, offering a rewarding space to make relationships, participate in self-care, share in leisure, and build potential livelihoods, with entertainment becoming a pleasing secondary function.
  • Item
    Thumbnail Image
    Designing a tangible user interface for the learning of motor skills in spinal mobilisation
    Chacon Salas, Dimas Antony ( 2018)
    Current techniques in the learning of psychomotor skills in physiotherapy, especially in spinal mobilisation, follow the traditional classroom approach: an expert performs a demonstration and students try to emulate the task by practising on each other while receiving mostly verbal feedback from the instructor. The introduction of a tailored tangible user interface would overcome the limitation of requiring the presence of a tutor and an extra fellow student, improving the scalability of the teaching delivery, and provide more objective feedback. Inspired by this opportunity, this work presents SpinalLog, a visuo-haptic interface that replicates the shape and deformable sensation of a human lower spine for the learning of spinal mobilisation techniques by employing conductive foam. This smart material is used simultaneously to sense vertebral displacements and provide passive haptic feedback to the user, emulating the flexibility of a spine. However, there is a need to understand the impact of the feedback provided in the learning of spinal mobilisation. Therefore, this work aims to design and implement SpinalLog to improve the teaching of this activity, and to investigate the effect of visual feedback, deformable haptic perception and shape fidelity in the learning of this delicate psychomotor task. We evaluated each of these three features—Visual Feedback, Passive Haptic Feedback, and Physical Fidelity—in the first part of an experiment to understand their effects on physiotherapy students' ability to replicate a mobilisation pattern recorded by an expert. Whereas in the second and last part of the experiment we presented the full features of our system to the students to gather their viewpoint for future improvement. From the first part of the experiment, we found that simultaneous feedback has the largest effect, followed by passive haptic feedback. The high fidelity of the interface has little quantitative effect, but it plays an important role in students' perceptions of the benefit of the system. From the second part of the experiment, we found that students had a favourable view on the SpinalLog suggesting improvements for the shape fidelity and the visual components.