With the recent popularity of electronic medical records, enormous amount of medical data is being generated every day at an exponential rate.Machine learning methods have been shown in many studies to be capable of producing automatic medical diagnostic models such as automated prognostic models. However, many powerful machine learning algorithms such as support vector machine (SVM), Random Forest (RF) or Kernel Logistic Regression (KLR) are unbearably slow for very large datasets. This makes their use in medical research limited to small to medium scale problems.This study is motivated by an ongoing research on prostate cancer mortality prediction for a national representative of US population where the SVM and RF took several hours or days to trainwhereas simple linear methods such as logistic regression or linear discriminant analysis take minutes or even seconds.Because, most real-world problems are non-linear, this paper presents a large scale algorithm enabling a recently proposed least squares extreme logistic regression to learn very large datasets. The algorithm is shown on a case study of mortality prediction for men diagnosed with early stage prostate cancer to provide very fast and more accurate result than standard statistical methods.

This work investigates how wireless signal attenuation affects a team of mobile robots performing exploration. Many coverage tasks, such as search and rescue or exploration, can be performed more effectively when robots communicate with one other. However, in real world environments, maintaining communication can be difficult due to unpredictable wireless signal propagation. In this paper, we investigate the wireless network connectivity in an outdoor area that consist of concrete wall and pillars. Preliminary simulation and physical experimental results are compared to demonstrate the effects of signal attenuation on robot team performance.

Probabilistic Relational Models (PRMs) are directed probabilistic graphical models representing a factored joint distribution over a set of random variables for relational datasets.While regular PRMs define probabilistic dependencies between classes’ descriptive attributes, an extension called PRM with Reference Uncertainty (PRM-RU) allows in addition to manage link uncertainty between them, by adding random variables called selectors. In order to avoid variables with large domains, selectors are associated with partition functions,mapping objects to a set of clusters, and selectors’ distributions are defined over the set of clusters. In PRM-RU, the definition of partition functions constrains us to learn them only from concerned individuals entity attributes and to assign the same cluster to a pair of individuals having the same attributes values. This constraint is actually based on a strong assumption which is not generalizable and can lead to an under usage of relationship data for learning. For these reasons,we relax this constraint in this paper and propose a different partition function learning approach based on relationship data clustering. We empirically show that this approach provides better results than attribute-based learning in the case where relationship topology is independent from involved entity attributes values, and that it gives close results whenever the attributes assumption is correct.

In this paper we propose a novel hierarchical multi-label clas- sification approach for tree and directed acyclic graph (DAG) hierarchies. The method predicts a single path (from the root to a leaf node) for tree hierarchies, and multiple paths for DAG hierarchies, by combining the predictions of every node in each possible path. In contrast with previous approaches, we evaluate all the paths, training local classifiers for each non-leaf node. The approach incorporates two contributions; (i) a cost is assigned to each node depending on the level it has in the hierarchy, giving more weight to correct predic- tions at the top levels; (ii) the relations between the nodes in the hierarchy are considered, by incorporating the parent label as in chained classifiers. The proposed approach was experimentally evaluated with 10 tree and 8 DAG hierarchi- cal datasets in the domain of protein function prediction. It was contrasted with various state-of-the-art hierarchical clas- sifiers using four common evaluation measures. The results show that our method is superior in almost all measures, and this difference is more significant in the case of DAG struc- tures.

Over the analyses being increasingly diversified, the notion of referential becomes more and more important in the theoretical plan and in the associated descriptions. We introduce the specification of this notion of referential with several examples, and then we propose its formalized approach which would admit the computing treatments later. We put this notion within the theoretical framework of the temporality (i.e. aspecto-temporal theory) apprehended by the languages while arguing its introduction and formalizing their concepts.

Ordinal conditional functions (OCFs) provide a semantic domain for qualitative conditionals of the form "if A, then (normally) B" by ordering worlds according to their degree of surprise. Transferring the idea of maximum entropy to a more qualitative domain, c-representations of a knowledge base R consisting of a set of conditionals have been defined as OCFs satisfying in particular the property of conditional indifference. While c-representations for R can be specified as the solutions of a constraint satisfaction problem CR(R), it has been an open problem whether there may be different minimal c-representations induced by minimal solutions of CR(R). Another open question has been whether particular inequations in CR(R) may be sharpened by transforming them into equations without loosing any minimal solutions, taking different notions of minimality into account. In this paper, we answer both questions and discuss further aspects of OCF minimality.

We describe a collaborative project between our research group and a small west-coast business to apply machine learning techniques to a document processing task. This experience suggests two key points: (1) even as machine learning and artificial intelligence matures, there are many business applications that have not yet exploited these techniques; and (2) academically well-established machine learning techniques have much to offer both in terms of flexibility and economic benefit.

Ideal portfolio creation has been the focus of consider- able machine learning research in the domain of finance. In this paper, the development of a two-stage platform for generating stable stock-based portfolios is explored. The first stage involves clustering of stocks based on time-weighted correlations, using a modified version of the K-Means++ algorithm. This clustering helps in the quantification of portfolio diversification at a later stage. In the second step, a genetic paradigm is employed to optimize the returns of the portfolio in such a way as to ensure its diversification at the same time. This leads to the formation of a portfolio that shows a high and sta- ble Markowitz ratio of returns/risk. The experimental results support the central hypothesis, and hint at possi- ble commercial applications.

The track on applied natural language processing (ANLP) is a forum for researchers working in natural language processing (NLP), computational linguistics (CL), and related areas.

Bidirectional Associative Memories (BAMs) are artificial neural networks that can learn and recall various types of associations. Although BAMs have shown great promise at modeling human cognitive processes, these models have often been investigated under optimal conditions in which the network is fully connected. Whereas some BAM models have shown to be robust to connection sparseness, those particular models could not handle highly sparse connectivity, unlike the human brain. This paper shows that a particular type of BAM can perform learning and recall under higher levels of sparse connectivity by increasing input dimensionality. This study provides a better understanding of the conditions impacting the convergence of the learning in BAM models and introduces a new avenue of research in learning in biological levels of sparseness, namely network dimensionality.