Catálogo de publicaciones - libros

This paper proposes a new nonlinear dimensionality reduction algorithm called RCTKE for highly structured data. It is built on the original TKE by incorporating a mapping function into the objective functional of TKE as regularization terms where the mapping function can be learned from training data and be used for novel samples. The experimental results on highly structured data is used to verify the effectiveness of the algorithm.

In this paper, we propose a method where humanoid robot acquires meanings of grammatical connectives using direct physical feedback from human. Our system acquired 70% connectives of all connectives taught by subjects. It can be also said that robot partially learned the concept of time.

A key challenge in data stream classification is to detect changes of the concept underlying the data, and accurately and efficiently adapt classifiers to each concept change. Most existing methods for handling concept changes take a windowing approach, where only recent instances are used to update classifiers while old instances are discarded indiscriminately. However this approach can often be undesirably aggressive because many old instances may not be affected by the concept change and hence can contribute to training the classifier, for instance, reducing the classification variance error caused by insufficient training data. Accordingly this paper proposes a cellular automata (CA) approach that feeds classifiers with most instead of most instances. The strength of CA is that it breaks a complicated process down into smaller adaptation tasks, for each a single automaton is responsible. Using neighborhood rules embedded in each automaton and emerging time of instances, this approach assigns a relevance weight to each instance. Instances with high enough weights are selected to update classifiers. Theoretical analyses and experimental results suggest that a good choice of local rules for CA can help considerably speed up updating classifiers corresponding to concept changes, increase classifiers’ robustness to noise, and thus offer faster and better classifications for data streams.

In this paper, we utilize a combination of SWEBOK and text categorization to categorize software engineering knowledge. SWEBOK serves as a backbone taxonomy while text categorization provides a collection of algorithms including knowledge representation, feature enrichment and machine learning. Firstly, fundamental knowledge types in software engineering are carefully analyzed as well as their characteristics. Then, incorporated with SWEBOK, we propose a knowledge categorization methodology as well as its implementing algorithms. Finally, we conduct experiments to evaluate the proposed method. The experimental results demonstrate that our methodology can serve as an effective solution for the categorization of software engineering knowledge.

To navigate successfully, a mobile robot must be able to estimate the spatial relationships of the objects of interest in its environment accurately. The main advantage of a bearing-only (SLAM) system is that it requires only a cheap vision sensor to enable a mobile robot to gain knowledge of its environment and navigate. In this paper, we focus on the representation of the spatial uncertainty of landmarks caused by sensor noise. We follow a principled approach for computing the (PDFs) of landmark positions when an initial observation is made. We characterize the PDF (,) of a landmark position expressed in polar coordinates when and are independent, and the marginal probability () of the PDF is constrained to be uniform.

Many real-world problems are over-constrained and require search techniques adapted to optimising cost functions rather than searching for consistency. This makes the MAX-SAT problem an important area of research for the satisfiability (SAT) community. In this study we perform an empirical analysis of several of the best performing SAT local search techniques in the domain of unweighted MAX-SAT. In particular, we test two of the most recently developed SAT clause weight redistribution algorithms, DDFW and DDFW, against three more well-known techniques (RSAPS, AdaptNovelty and PAWS). Based on an empirical study across a range of previously studied problems we conclude that DDFW is the most promising algorithm in terms of robust average performance.

Population Projection is the numerical outcome of a specific set of assumptions about future population changes. It is indispensable to the planning of sites as almost all successive planning activities such as the identification of land and housing supply, the release of land, the planning and construction of social and physical infrastructure are population related. This paper proposes a new hierarchical framework based on Hidden Markov Model (HMM), called HMM-Bin framework, for use in long-term population projection. Analyses of various existing suburbs indicate it outperforms traditional Cohort Component model and simple HMM in terms of less data dependency, output flexibility and long-term projection accuracy.

PMP is a well-known NP-hard problem with extensively wide applications in location science and clustering. In this paper, we presented computational complexity results about the backbone, the shared common parts of all the optimal solutions to the PMP. We showed that it is intractable to approximate the backbone of the PMP with any performance guarantee under the assumption that  ≠ .

In this paper, we present a novel associative memory approach for pattern recognition termed as Distributed Hierarchical Graph Neuron (DHGN). DHGN is a scalable, distributed, and one-shot learning pattern recognition algorithm which uses graph representations for pattern matching without increasing the computation complexity of the algorithm. We have successfully tested this algorithm for character patterns with structural and random distortions. The pattern recognition process is completed in one-shot and within a fixed number of steps.

Typically association rule mining only considers positive frequent itemsets in rule generation, where rules involving only the presence of items are generated. In this paper we consider the complementary problem of negative association rule mining, which generates rules describing the absence of itemsets from transactions. We describe a new approach called MINR (Mining Interesting Negative Rules) to efficiently find all interesting negative association rules. Here we only consider the presence or absence of itemsets that are strongly associated. Our approach does not require a user defined support threshold, and is based on pruning items that occur together by coincidence. For every individual itemset we calculate two custom thresholds based on their support: the positive and negative chance thresholds. We compared our implementation against Pearson correlation.