No description available.

Domain-independent feature learning is a hard problem. This is reflected by lack of broad research in the area. The goal of General Game Playing (GGP) can be described as designing computer programs that can play a variety of games given only a logical game description. Any learning has to be domain-independent in the GGP framework. Learning algorithms have not been an essential part of all successful GGP programs. This thesis presents a feature learning approach, GIFL, for 2-player, alternating move games using state differences. The algorithm is simple, robust and improves the quality of play. GIFL is implemented in a GGP program, Maligne. The experiments show that GIFL outperforms standard UCT algorithm in nine out of fifteen games and loses performance only in one game.

Learning discriminative feature representations has attracted a great deal of attention since it is a critical step to facilitate the subsequent classification, retrieval and recommendation tasks. In this dissertation, besides incorporating prior knowledge about image labels into the image classification as most prevalent feature learning methods currently do, we also explore some other general-purpose priors and verify their effectiveness in the discriminant feature learning. As a more powerful representation can be learned by implementing such general priors, our approaches achieve state-of-the-art results on challenging benchmarks. We elaborate on these general-purpose priors and highlight where we have made novel contributions. We apply sparsity and hierarchical priors to the explanatory factors that describe the data, in order to better discover the data structure. More specifically, in the first approach we propose that we only incorporate sparse priors into the feature learning. To this end, we present a support discrimination dictionary learning method, which finds a dictionary under which the feature representation of images from the same class have a common sparse structure while the size of the overlapped signal support of different classes is minimised. Then we incorporate sparse priors and hierarchical priors into a unified framework, that is capable of controlling the sparsity of the neuron activation in deep neural networks. Our proposed approach automatically selects the most useful low-level features and effectively combines them into more powerful and discriminative features for our specific image classification problem. We also explore priors on the relationships between multiple factors. When multiple independent factors exist in the image generation process and only some of them are of interest to us, we propose a novel multi-task adversarial network to learn a disentangled feature which is optimized with respect to the factor of interest to us, while being distraction factors agnostic. When common factors exist in multiple tasks, leveraging common factors cannot only make the learned feature representation more robust, but also enable the model to generalise from very few labelled samples. More specifically, we address the domain adaptation problem and propose the re-weighted adversarial adaptation network to reduce the feature distribution divergence and adapt the classifier from source to target domains.

Current unsupervised approaches for keyphrase extraction compute a single importance score for each candidate word by considering the number and quality of its associated words in the graph and they are not flexible enough to incorporate multiple types of information. For instance, nodes in a network may exhibit diverse connectivity patterns which are not captured by the graph-based ranking methods. To address this, we present a new approach to keyphrase extraction that represents the document as a word graph and exploits its structure in order to reveal underlying explanatory factors hidden in the data that may distinguish keyphrases from non-keyphrases. Experimental results show that our model, which uses phrase graph representations in a supervised probabilistic framework, obtains remarkable improvements in performance over previous supervised and unsupervised keyphrase extraction systems.

In this paper, we introduce a novel set of features for robust object recognition, which exhibits outstanding performances on a variety ofobject categories while being capable of learning from only a fewtraining examples. Each element of this set is a complex featureobtained by combining position- and scale-tolerant edge-detectors overneighboring positions and multiple orientations.Our system - motivated by a quantitative model of visual cortex -outperforms state-of-the-art systems on a variety of object imagedatasets from different groups. We also show that our system is ableto learn from very few examples with no prior category knowledge. Thesuccess of the approach is also a suggestive plausibility proof for aclass of feed-forward models of object recognition in cortex. Finally,we conjecture the existence of a universal overcompletedictionary of features that could handle the recognition of all objectcategories.

京都大学 / 0048 / 新制・課程博士 / 博士(情報学) / 甲第19111号 / 情博第557号 / 新制||情||98(附属図書館) / 32062 / 京都大学大学院情報学研究科知能情報学専攻 / (主査)教授 松山 隆司, 教授 美濃 導彦, 准教授 梁 雪峰 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

Information mining and knowledge learning from sequential data is a field of growing importance in both industrial and academic fields. Sequential data, which is the natural representation format of the information flow in many applications, usually carries enormous information and is able to help researchers gain insights for various tasks such as airport threat detection, cyber-attack detection, recommender system, point-of-interest (POI) prediction, and citation forecasting. This dissertation focuses on developing the methods for sequential data-driven applications and evolutionary dynamics characterization for various topics such as transit service disruption detection, early event detection on social media, technology opportunity discovery, and traffic incident impact analysis. In particular, four specific applications are studied with four proposed novel methods, including a spatiotemporal feature learning framework for transit service disruption detection, a multi-task learning framework for cybersecurity event detection, citation dynamics modeling via multi-context attentional recurrent neural networks, and traffic incident impact forecasting via hierarchical spatiotemporal graph neural networks. For the first of these methods, the existing transit service disruption detection methods usually suffer from two significant shortcomings: 1) failing to modulate the sparsity of the social media feature domain, i.e., only a few important ``particles'' are indeed related to service disruption among the massive volume of data generated every day and 2) ignoring the real-world geographical connections of transit networks as well as the semantic consistency existing in the problem space. This work makes three contributions: 1) developing a spatiotemporal learning framework for metro disruption detection using open-source data, 2) modeling semantic similarity and spatial connectivity among metro lines in feature space, and 3) developing an optimization algorithm for solving the multi-convex and non-smooth objective function efficiently. For the second of these methods, the conventional studies in cybersecurity detection suffer from the following shortcomings: 1) unable to capture weak signals generated by the cyber-attacks on small organizations or individual accounts, 2) lack of generalization of distinct types of security incidents, and 3) failing to consider the relatedness across different types of cyber-attacks in the feature domain. Three contributions are made in this work: 1) formulating the problem of social media-based cyber-attack detection into the multi-task learning framework, 2) modeling multi-type task relatedness in feature space, and 3) developing an efficient algorithm to solve the non-smooth model with inequality constraints. For the third of these methods, conventional citation forecasting methods are using the traditional temporal point process, which suffers from several drawbacks: 1) unable to predict the technological categories of citing documents and thus are incapable of technological diversity assessment, and 2) require prior domain knowledge and thus are hard to extend to different research areas. Two contributions are made in this work: 1) formulating a novel framework to provide long-term citation predictions in an end-to-end fashion by integrating the process of learning intensity function representations and the process of predicting future citations and 2) designing two novel temporal attention mechanisms to improve the model's ability to modulate complicated temporal dependencies and to allow the model to dynamically combine the observation and prediction sides during the learning process. For the fourth of these methods, the previous work treats the traffic sensor readings as the features and views the incident duration prediction as a feature-driven regression, which typically suffers from three drawbacks: 1) ignoring the existence of the road-sensor hierarchical structure in the real-world traffic network, 2) unable to learn and modulate the hidden temporal patterns in the sensor readings, and 3) lack of consideration of the spatial connectivity between arterial roads and traffic sensors. This work makes three significant contributions: 1) designing a hierarchical graph convolutional network architecture for modeling the road-sensor hierarchy, 2) proposing novel spatiotemporal attention mechanism on the sensor- and road-level features for representation learning, and 3) presenting a graph convolutional network-based method for incident representation learning via spatial connectivity modeling and traffic characteristics modulation. / Doctor of Philosophy / Information mining and knowledge learning from sequential data is a field of growing importance in both industrial and academic fields. Sequential data, which is the natural representation format of the information flow in many applications, usually carries enormous information and is able to help researchers gain insights for various tasks such as airport threat detection, cyber-attack detection, recommender system, point-of-interest (POI) prediction, and citation forecasting. This dissertation focuses on developing the methods for sequential data-driven applications and evolutionary dynamics characterization for various topics such as transit service disruption detection, early event detection on social media, technology opportunity discovery, and traffic incident impact analysis. In particular, four specific applications are studied with four proposed novel methods, including a spatiotemporal feature learning framework for transit service disruption detection, a multi-task learning framework for cybersecurity event detection, citation dynamics modeling via multi-context attentional recurrent neural networks, and traffic incident impact forecasting via hierarchical spatiotemporal graph neural networks. For the first of these methods, the existing transit service disruption detection methods usually suffer from two significant shortcomings: 1) failing to modulate the sparsity of the social media feature domain, i.e., only a few important ``particles'' are indeed related to service disruption among the massive volume of data generated every day and 2) ignoring the real-world geographical connections of transit networks as well as the semantic consistency existing in the problem space. This work makes three contributions: 1) developing a spatiotemporal learning framework for metro disruption detection using open-source data, 2) modeling semantic similarity and spatial connectivity among metro lines in feature space, and 3) developing an optimization algorithm for solving the multi-convex and non-smooth objective function efficiently. For the second of these methods, the conventional studies in cybersecurity detection suffer from the following shortcomings: 1) unable to capture weak signals generated by the cyber-attacks on small organizations or individual accounts, 2) lack of generalization of distinct types of security incidents, and 3) failing to consider the relatedness across different types of cyber-attacks in the feature domain. Three contributions are made in this work: 1) formulating the problem of social media-based cyber-attack detection into the multi-task learning framework, 2) modeling multi-type task relatedness in feature space, and 3) developing an efficient algorithm to solve the non-smooth model with inequality constraints. For the third of these methods, conventional citation forecasting methods are using the traditional temporal point process, which suffers from several drawbacks: 1) unable to predict the technological categories of citing documents and thus are incapable of technological diversity assessment, and 2) require prior domain knowledge and thus are hard to extend to different research areas. Two contributions are made in this work: 1) formulating a novel framework to provide long-term citation predictions in an end-to-end fashion by integrating the process of learning intensity function representations and the process of predicting future citations and 2) designing two novel temporal attention mechanisms to improve the model's ability to modulate complicated temporal dependencies and to allow the model to dynamically combine the observation and prediction sides during the learning process. For the fourth of these methods, the previous work treats the traffic sensor readings as the features and views the incident duration prediction as a feature-driven regression, which typically suffers from three drawbacks: 1) ignoring the existence of the road-sensor hierarchical structure in the real-world traffic network, 2) unable to learn and modulate the hidden temporal patterns in the sensor readings, and 3) lack of consideration of the spatial connectivity between arterial roads and traffic sensors. This work makes three significant contributions: 1) designing a hierarchical graph convolutional network architecture for modeling the road-sensor hierarchy, 2) proposing novel spatiotemporal attention mechanism on the sensor- and road-level features for representation learning, and 3) presenting a graph convolutional network-based method for incident representation learning via spatial connectivity modeling and traffic characteristics modulation.

Yes / Gait is a unique non-invasive biometric form that can be utilized to effectively recognize persons, even when they prove to be uncooperative. Computer-aided gait recognition systems usually use image sequences without considering covariates like clothing and possessions of carrier bags whilst on the move. Similarly, in gait recognition, there may exist unknown covariate conditions that may affect the training and testing conditions for a given individual. Consequently, common techniques for gait recognition and measurement require a degree of intervention leading to the introduction of unknown covariate conditions, and hence this significantly limits the practical use of the present gait recognition and analysis systems. To overcome these key issues, we propose a method of gait analysis accounting for both known and unknown covariate conditions. For this purpose, we propose two methods, i.e., a Convolutional Neural Network (CNN) based gait recognition and a discriminative features-based classification method for unknown covariate conditions. The first method can handle known covariate conditions efficiently while the second method focuses on identifying and selecting unique covariate invariant features from the gallery and probe sequences. The feature set utilized here includes Local Binary Patterns (LBP), Histogram of Oriented Gradients (HOG), and Haralick texture features. Furthermore, we utilize the Fisher Linear Discriminant Analysis for dimensionality reduction and selecting the most discriminant features. Three classifiers, namely Random Forest, Support Vector Machine (SVM), and Multilayer Perceptron are used for gait recognition under strict unknown covariate conditions. We evaluated our results using CASIA and OUR-ISIR datasets for both clothing and speed variations. As a result, we report that on average we obtain an accuracy of 90.32% for the CASIA dataset with unknown covariates and similarly performed excellently on the ISIR dataset. Therefore, our proposed method outperforms existing methods for gait recognition under known and unknown covariate conditions. / This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (NRF-2019R1F1A1060668).

Various hand-crafted features have been explored for medical image classification, which include SIFT and Local Binary Patterns (LBP). However, hand-crafted features may not be optimally discriminative for classifying images from particular domains (e.g. colonoscopy), as not necessarily tuned to the domain’s characteristics. In this work, I give emphasis on learning highly discriminative local features and image representations to achieve the best possible classification performance for medical images, particularly for colonoscopy and histology (cell) images. I propose approaches to learn local features using unsupervised and weakly-supervised methods, and an approach to improve the feature encoding methods such as bag-of-words. Unlike the existing work, the proposed weakly-supervised approach uses image-level labels to learn the local features. Requiring image-labels instead of region-level labels makes annotations less expensive, and closer to the data normally available from normal clinical practice, hence more feasible in practice. In this thesis, first, I propose a generalised version of the LBP descriptor called the Generalised Local Ternary Patterns (gLTP), which is inspired by the success of LBP and its variants for colonoscopy image classification. gLTP is robust to both noise and illumination changes, and I demonstrate its competitive performance compared to the best performing LBP-based descriptors on two different datasets (colonoscopy and histology). However LBP-based descriptors (including gLTP) lose information due to the binarisation step involved in their construction. Therefore, I then propose a descriptor called the Extended Multi-Resolution Local Patterns (xMRLP), which is real-valued and reduces information loss. I propose unsupervised and weakly-supervised learning approaches to learn the set of parameters in xMRLP. I show that the learned descriptors give competitive or better performance compared to other descriptors such as root-SIFT and Random Projections. Finally, I propose an approach to improve feature encoding methods. The approach captures inter-cluster features, providing context information in the feature as well as in the image spaces, in addition to the intra-cluster features often captured by conventional feature encoding approaches. The proposed approaches have been evaluated on three datasets, 2-class colonoscopy (2, 100 images), 3-class colonoscopy (2, 800 images) and histology (public dataset, containing 13, 596 images). Some experiments on radiology images (IRMA dataset, public) also were given. I show state-of-the-art or superior classification performance on colonoscopy and histology datasets.

Scene perception is a fundamental aspect of vision. Humans are capable of analyzing behaviorally-relevant scene properties such as spatial layouts or scene categories very quickly, even from low resolution versions of scenes. Although humans perform these tasks effortlessly, they are very challenging for machines. Developing methods that well capture the properties of the representation used by the visual system will be useful for building computational models that are more consistent with perception. While it is common to use hand-engineered features that extract information from predefined dimensions, they require careful tuning of parameters and do not generalize well to other tasks or larger datasets. This thesis is driven by the hypothesis that the perceptual representations are adapted to the statistical properties of natural visual scenes. For developing statistical features for global-scale structures (low spatial frequency information that encompasses entire scenes), I propose to train hierarchical probabilistic models on whole scene images. I first investigate statistical clusters of scene images by training a mixture model under the assumption that each image can be decoded by sparse and independent coefficients. Each cluster discovered by the unsupervised classifier is consistent with the high-level semantic categories (such as indoor, outdoor-natural and outdoor-manmade) as well as perceptual layout properties (mean depth, openness and perspective). To address the limitation of mixture models in their assumptions of a discrete number of underlying clusters, I further investigate a continuous representation for the distributions of whole scenes. The model parameters optimized for natural visual scenes reveal a compact representation that encodes their global-scale structures. I develop a probabilistic similarity measure based on the model and demonstrate its consistency with the perceptual similarities. Lastly, to learn the representations that better encode the manifold structures in general high-dimensional image space, I develop the image normalization process to find a set of canonical images that anchors the probabilistic distributions around the real data manifolds. The canonical images are employed as the centers of the conditional multivariate Gaussian distributions. This approach allows to learn more detailed structures of the local manifolds resulting in improved representation of the high level properties of scene images.