Induction by a Hilbert hypercube representation

This thesis describes a novel connectionist machine utilizing induction by a Hilbert hypercube representation. This representation offers a number of distinct advantages which are described. We construct a theoretical and practical learning machine which lies in an area of overlap between three disciplines - neural nets, machine learning and knowledge acquisition - hence it is refered to as a "coalesced" machine. To this unifying aspect is added the various advantages of its orthogonal lattice structure as against less structured nets. We discuss the case for such a fundamental and low level empirical learning tool and the assumptions behind the machine are clearly outlined. Our theory of an orthogonal lattice structure the Hilbert hypercube of an n-dimensional space using a complemented distributed lattice as a basis for supervised learning is derived from first principles on clearly laid out scientific principles. The resulting "subhypercube theory" was implemented in a development machine which was then used to test the theoretical predictions again under strict scientific guidelines. The scope, advantages and limitations of this machine were tested in a series of experiments. Novel and seminal properties of the machine include: the "metrical", deterministic and global nature of its search; complete convergence invariably producing minimum polynomial solutions for both disjuncts and conjuncts even with moderate levels of noise present; a learning engine which is mathematically analysable in depth based upon the "complexity range" of the function concerned; a strong bias towards the simplest possible globally (rather than locally) derived "balanced" explanation of the data; the ability to cope with variables in the network; and new ways of reducing the exponential explosion. Performance issues were addressed and comparative studies with other learning machines indicates that our novel approach has definite value and should be further researched.

Learning task specific distributed paragraph representations using a 2-tier convolutional neural network

We introduce a type of 2-tier convolutional neural network model for learning distributed paragraph representations for a special task (e.g. paragraph or short document level sentiment analysis and text topic categorization). We decompose the paragraph semantics into 3 cascaded constitutes: word representation, sentence composition and document composition. Specifically, we learn distributed word representations by a continuous bag-of-words model from a large unstructured text corpus. Then, using these word representations as pre-trained vectors, distributed task specific sentence representations are learned from a sentence level corpus with task-specific labels by the first tier of our model. Using these sentence representations as distributed paragraph representation vectors, distributed paragraph representations are learned from a paragraph-level corpus by the second tier of our model. It is evaluated on DBpedia ontology classification dataset and Amazon review dataset. Empirical results show the effectiveness of our proposed learning model for generating distributed paragraph representations.

Building belief in organizational learning: the role of organizational learning efficacy

There appears to be a missing dimension in OL literature to embrace the collective experience of emotion, both within groups and communities and also across the organization as a whole. The concept of OL efficacy- as a stimulus offering energy and direction for learning - remains unexplored. This research involved engaging with a company we have called ‘Electroco’ in depth to create a rich and nuanced representation of OL and members’ perceptions of OL over an extended time-frame (five years). We drew upon grounded theory research methodology (Locke, 2001), to elicit feedback from the organization, which was then used to inform future research plans and/ or refine emerging ideas. The concept of OL efficacy gradually emerged as a factor to be considered when exploring the relationship between individual learning and OL. . Bearing in mind Bandura’s (1982) conceptualization of self-efficacy (linked with mastery, modelling, verbal persuasion and emotional arousal), we developed a coding strategy encompassing these four factors as conceptualized at the organizational level. We added a fifth factor: ‘control of OL.’ We focused on feelings across the organization and the extent of consensus or otherwise around these five attributes. The construct has potential significance for how people are managed in many ways. Not only is OL efficacy is difficult for competitors to copy (arising as it does from the collective experience of working within a specific context); the self-efficacy concept suggests that success can be engineered with ‘small wins’ to reinforce mastery perceptions. Leaders can signal the importance of interaction with the external context, and encourage reflection on the strategies adopted by competitors or benchmark organizations (modelling). The theory also underlines the key role managers may play in persuading others about their organization’s propensity to learn (by focusing on success stories, for example). Research is set to continue within other sectors, including the high-performance financial service sector as well as the health-care technology sector.

Increased salience of gains versus decreased associative learning differentiate bipolar disorder from schizophrenia during incentive decision making

Background Abnormalities in incentive decision making, typically assessed using the Iowa Gambling Task (IGT), have been reported in both schizophrenia (SZ) and bipolar disorder (BD). We applied the Expectancy-Valence (E-V) model to determine whether motivational, cognitive and response selection component processes of IGT performance are differentially affected in SZ and BD. Method Performance on the IGT was assessed in 280 individuals comprising 70 remitted patients with SZ, 70 remitted patients with BD and 140 age-, sex-and IQ-matched healthy individuals. Based on the E-V model, we extracted three parameters, 'attention to gains or loses', 'expectancy learning' and 'response consistency', that respectively reflect motivational, cognitive and response selection influences on IGT performance. Results Both patient groups underperformed in the IGT compared to healthy individuals. However, the source of these deficits was diagnosis specific. Associative learning underlying the representation of expectancies was disrupted in SZ whereas BD was associated with increased incentive salience of gains. These findings were not attributable to non-specific effects of sex, IQ, psychopathology or medication. Conclusions Our results point to dissociable processes underlying abnormal incentive decision making in BD and SZ that could potentially be mapped to different neural circuits. © 2012 Cambridge University Press.

Semi-supervised learning of statistical models for natural language understanding

Natural language understanding is to specify a computational model that maps sentences to their semantic mean representation. In this paper, we propose a novel framework to train the statistical models without using expensive fully annotated data. In particular, the input of our framework is a set of sentences labeled with abstract semantic annotations. These annotations encode the underlying embedded semantic structural relations without explicit word/semantic tag alignment. The proposed framework can automatically induce derivation rules that map sentences to their semantic meaning representations. The learning framework is applied on two statistical models, the conditional random fields (CRFs) and the hidden Markov support vector machines (HM-SVMs). Our experimental results on the DARPA communicator data show that both CRFs and HM-SVMs outperform the baseline approach, previously proposed hidden vector state (HVS) model which is also trained on abstract semantic annotations. In addition, the proposed framework shows superior performance than two other baseline approaches, a hybrid framework combining HVS and HM-SVMs and discriminative training of HVS, with a relative error reduction rate of about 25% and 15% being achieved in F-measure.

Production-frame Model of Representation of Data Domain

One of the most important problems of e-learning system is studied in given paper. This problem is building of data domain model. Data domain model is based on usage of correct organizing knowledge base. In this paper production-frame model is offered, which allows structuring data domain and building flexible and understandable inference system, residing in production system.

A Domain-Independent Otology-based Approach to Representation of Courseware Knowledge

This paper proposes an ontology-based approach to representation of courseware knowledge in different domains. The focus is on a three-level semantic graph, modeling respectively the course as a whole, its structure, and domain contents itself. The authors plan to use this representation for flexibie e- learning and generation of different study plans for the learners.

Improving distributed representation of word sense via WordNet Gloss composition and context clustering

In recent years, there has been an increas-ing interest in learning a distributed rep-resentation of word sense. Traditional context clustering based models usually require careful tuning of model parame-ters, and typically perform worse on infre-quent word senses. This paper presents a novel approach which addresses these lim-itations by first initializing the word sense embeddings through learning sentence-level embeddings from WordNet glosses using a convolutional neural networks. The initialized word sense embeddings are used by a context clustering based model to generate the distributed representations of word senses. Our learned represen-tations outperform the publicly available embeddings on 2 out of 4 metrics in the word similarity task, and 6 out of 13 sub tasks in the analogical reasoning task.

A Multi-View Learning Algorithm

Report published in the Proceedings of the National Conference on "Education and Research in the Information Society", Plovdiv, May, 2014

A gloss composition and context clustering based distributed word sense representation model

In recent years, there has been an increasing interest in learning a distributed representation of word sense. Traditional context clustering based models usually require careful tuning of model parameters, and typically perform worse on infrequent word senses. This paper presents a novel approach which addresses these limitations by first initializing the word sense embeddings through learning sentence-level embeddings from WordNet glosses using a convolutional neural networks. The initialized word sense embeddings are used by a context clustering based model to generate the distributed representations of word senses. Our learned representations outperform the publicly available embeddings on half of the metrics in the word similarity task, 6 out of 13 sub tasks in the analogical reasoning task, and gives the best overall accuracy in the word sense effect classification task, which shows the effectiveness of our proposed distributed distribution learning model.

A study of the relationship between educational placement and the achievement of urban low socioeconomic Hispanic middle school students with and without specific learning disabilities

Public schools traditionally have been held accountable for educating the majority of the nation’s school children, and through the years, these schools have been evaluated in a variety of ways. Currently, evaluation measures for accountability purposes consist solely of standardized test scores. In the past, only test scores of general education students were analyzed. Laws governing the education of students with disabilities, however, have extended accountability measures not only to include those students, but to report their scores in a disaggregated form (No Child Left Behind Act, 2001). The recent emphasis on accountability and compliance has resulted in the need for schools to carefully examine how programs, services, and policies impact student achievement (Bowers & Figgers, 2003). ^ Standard-based school reform and accountability systems have raised expectations about student learning outcomes for all students, including those with disabilities and minority students. Yet, overall, racial/ethnic minority students are performing well below their White non-Hispanic peers in most academic areas. Additionally, with respect to special education, there exists an enduring problem of disproportionate representation of racial/ethnic minority students (National Research Council, 2000). ^ This study examined classroom placement (inclusive versus non-inclusive) relative to academic performance of urban, low socioeconomic Hispanic students with and without disabilities in secondary content area classrooms. A mixed method research design was used to investigate this important issue using data from a local school district and results from field observations. The study compared performance levels of four middle school Hispanic student subgroups (students with disabilities in inclusive settings, students without disabilities in inclusive settings, students with disabilities in resource settings, and student without disabilities in general education settings) each in their respective placements for two consecutive years, exploring existing practices within authentic settings. ^ Significant differences were found in the relationship of educational placement and achievement between grade level and disability in the areas of math and reading. Additionally, clear and important differences were observed in student-teacher interactions. Recommendations for further researchers and stakeholders include soliciting responses from teams at the schools composed of general education and special education teachers, administrative personnel, and students as well as broadening the study across grade levels and exceptionalities. ^

Algorithms for super-resolution of images based on Sparse Representation and Manifolds

lmage super-resolution is defined as a class of techniques that enhance the spatial resolution of images. Super-resolution methods can be subdivided in single and multi image methods. This thesis focuses on developing algorithms based on mathematical theories for single image super­ resolution problems. lndeed, in arder to estimate an output image, we adopta mixed approach: i.e., we use both a dictionary of patches with sparsity constraints (typical of learning-based methods) and regularization terms (typical of reconstruction-based methods). Although the existing methods already per- form well, they do not take into account the geometry of the data to: regularize the solution, cluster data samples (samples are often clustered using algorithms with the Euclidean distance as a dissimilarity metric), learn dictionaries (they are often learned using PCA or K-SVD). Thus, state-of-the-art methods still suffer from shortcomings. In this work, we proposed three new methods to overcome these deficiencies. First, we developed SE-ASDS (a structure tensor based regularization term) in arder to improve the sharpness of edges. SE-ASDS achieves much better results than many state-of-the- art algorithms. Then, we proposed AGNN and GOC algorithms for determining a local subset of training samples from which a good local model can be computed for recon- structing a given input test sample, where we take into account the underlying geometry of the data. AGNN and GOC methods outperform spectral clustering, soft clustering, and geodesic distance based subset selection in most settings. Next, we proposed aSOB strategy which takes into account the geometry of the data and the dictionary size. The aSOB strategy outperforms both PCA and PGA methods. Finally, we combine all our methods in a unique algorithm, named G2SR. Our proposed G2SR algorithm shows better visual and quantitative results when compared to the results of state-of-the-art methods.

Short and Long Term Representation of an Unfamiliar Tone Distribution

We report on a study conducted to extend our knowledge about the process of gaining a mental representation of music. Several studies, inspired by research on the statistical learning of language, have investigated statistical learning of sequential rules underlying tone sequences. Given that the mental representation of music correlates with distributional properties of music, we tested whether participants are able to abstract distributional information contained in tone sequences to form a mental representation. For this purpose, we created an unfamiliar music genre defined by an underlying tone distribution, to which 40 participants were exposed. Our stimuli allowed us to differentiate between sensitivity to the distributional properties contained in test stimuli and long term representation of the distributional properties of the music genre overall. Using a probe tone paradigm and a two-alternative forced choice discrimination task, we show that listeners are able to abstract distributional properties of music through mere exposure into a long term representation of music. This lends support to the idea that statistical learning is involved in the process of gaining musical knowledge.

Exploring the mobile structural assessment tool: Concept maps for learning website: Concept Maps for Learning Website

In this paper, we describe how the pathfinder algorithm converts relatedness ratings of concept pairs to concept maps; we also present how this algorithm has been used to develop the Concept Maps for Learning website (www.conceptmapsforlearning.com) based on the principles of effective formative assessment. The pathfinder networks, one of the network representation tools, claim to help more students memorize and recall the relations between concepts than spatial representation tools (such as Multi- Dimensional Scaling). Therefore, the pathfinder networks have been used in various studies on knowledge structures, including identifying students’ misconceptions. To accomplish this, each student’s knowledge map and the expert knowledge map are compared via the pathfinder software, and the differences between these maps are highlighted. After misconceptions are identified, the pathfinder software fails to provide any feedback on these misconceptions. To overcome this weakness, we have been developing a mobile-based concept mapping tool providing visual, textual and remedial feedback (ex. videos, website links and applets) on the concept relations. This information is then placed on the expert concept map, but not on the student’s concept map. Additionally, students are asked to note what they understand from given feedback, and given the opportunity to revise their knowledge maps after receiving various types of feedback.

An incremental learning algorithm for face recognition

[EN]In face recognition, where high-dimensional representation spaces are generally used, it is very important to take advantage of all the available information. In particular, many labelled facial images will be accumulated while the recognition system is functioning, and due to practical reasons some of them are often discarded. In this paper, we propose an algorithm for using this information. The algorithm has the fundamental characteristic of being incremental. On the other hand, the algorithm makes use of a combination of classification results for the images in the input sequence. Experiments with sequences obtained with a real person detection and tracking system allow us to analyze the performance of the algorithm, as well as its potential improvements.