Joint maximization of accuracy and information for learning the structure of a Bayesian network classifier

Machine Learning - Although recent studies have shown that a Bayesian network classifier (BNC) that maximizes the classification accuracy (i.e., minimizes the 0/1 loss function) is a powerful tool...     

The value of agreement a new boosting algorithm

In the past few years unlabeled examples and their potential advantage have received a lot of attention. In this paper a new boosting algorithm is presented where unlabeled examples are used to enforce agreement between several different learning algorithms. Not only do the learning algorithms learn from the given training set but they are supposed to do so while agreeing on the unlabeled examples. Similar ideas have been proposed before (for example, the Co-Training algorithm by Mitchell and Blum), but without a proof or under strong assumptions. In our setting, it is only assumed that all learning algorithms are equally adequate for the tasks. A new generalization bound is presented where the use of unlabeled examples results in a better ratio between training-set size and the resulting classifier's quality and thus reduce the number of labeled examples necessary for achieving it. The extent of this improvement depends on the diversity of the learners—a more diverse group of learners will result in a larger improvement whereas using two copies of a single algorithm gives no advantage at all. As a proof of concept, the algorithm, named Agreement Boost, is applied to two test problems. In both cases, using Agreement Boost results in an up to 40% reduction in the number of labeled examples.     

Approximate distributed top-<Emphasis Type="Italic">k</Emphasis> queries

We consider a distributed system where each node keeps a local count for items (similar to elections where nodes are ballot boxes and items are candidates). A top-k query in such a system asks which are the k items whose global count, across all nodes in the system, is the largest. In this paper, we present a Monte Carlo algorithm that outputs, with high probability, a set of k candidates which approximates the top-k items. The algorithm is motivated by sensor networks in that it focuses on reducing the individual communication complexity. In contrast to previous algorithms, the communication complexity depends only on the global scores and not on the partition of scores among nodes. If the number of nodes is large, our algorithm dramatically reduces the communication complexity when compared with deterministic algorithms. We show that the complexity of our algorithm is close to a lower bound on the cell-probe complexity of any non-interactive top-k approximation algorithm. We show that for some natural global distributions (such as the Geometric or Zipf distributions), our algorithm needs only polylogarithmic number of communication bits per node. An extended abstract of this paper appeared in Proc. 13th Int. Colloquium on Structural Information and Communication Complexity, SIROCCO 2006, Lecture Notes in Computer Science 4056, pp. 319–333.     

Online Scheduling with Interval Conflicts

In the problem of Scheduling with Interval Conflicts, there is a ground set of items indexed by integers, and the input is a collection of conflicts, each containing all the items whose index lies within some interval on the real line. Conflicts arrive in an online fashion. A scheduling algorithm must select, from each conflict, at most one survivor item, and the goal is to maximize the number (or weight) of items that survive all the conflicts they are involved in. We present a centralized deterministic online algorithm whose competitive ratio is O(lgσ), where σ is the size of the largest conflict. For the distributed setting, we present another deterministic algorithm whose competitive ratio is $2\left \lceil {\lg\sigma} \right \rceil $ in the special contiguous case, in which the item indices constitute a contiguous interval of integers. Our upper bounds are complemented by two lower bounds: one that shows that even in the contiguous case, all deterministic algorithms (centralized or distributed) have competitive ratio Ω(lgσ), and that in the non-contiguous case, no deterministic oblivious algorithm (i.e., a distributed algorithm that does not use communication) can have a bounded competitive ratio.     

General Perfectly Periodic Scheduling

In a perfectly periodic schedule, each job must be scheduled precisely every some fixed number of time units after its previous occurrence. Traditionally, motivated by centralized systems, the perfect periodicity requirement is relaxed, the main goal being to attain the requested average rate. Recently, motivated by mobile clients with limited power supply, perfect periodicity seems to be an attractive alternative that allows clients to save energy by reducing their "busy waiting" time. In this case, clients may be willing to compromise their requested service rate in order to get perfect periodicity. In this paper we study a general model of perfectly periodic schedules, where each job has a requested period and a length; we assume that m jobs can be served in parallel for some given m. Job lengths may not be truncated, but granted periods may be different than the requested periods. We present an algorithm which computes schedules such that the worst-case proportion between the requested period and the granted period is guaranteed to be close to the lower bound. This algorithm improves on previous algorithms for perfect schedules in providing a worst-case guarantee rather than an average-case guarantee, in generalizing unit length jobs to arbitrary length jobs, and in generalizing the single-server model to multiple servers.     

Efficient algorithms for center problems in cactus networks

Efficient algorithms for solving the center problems in weighted cactus networks are presented. In particular, we have proposed the following algorithms for the weighted cactus networks of size n$n$: an O(nlogn)$O(nlogn)$ time algorithm to solve the 1-center problem, and an O(nlog³n)$O\left(n{log}^{3}n\right)$ time algorithm to solve the weighted continuous 2-center problem. We have also provided improved solutions to the general p$p$-center problems in cactus networks. The developed ideas are then applied to solve the obnoxious 1-center problem in weighted cactus networks.     

Modelling Off-the-Shelf Information Systems Requirements: An Ontological Approach

Requirements for choosing off-the-shelf information systems (OISR) differ from requirements for development of new information systems in that they do not necessarily provide complete specifications, thus allowing flexibility in matching an existing IS to the stated needs. We present a framework for OISR conceptual models that consists of four essential elements: business processes, business rules, information objects and required system services. We formalise the definitions of these concepts based on an ontological model. The ontology-based OISR model provides a framework to evaluate modelling languages on how appropriate they are for OISR requirements specifications. The evaluation framework is applied to the Object-Process Methodology, and its results are compared with a similar evaluation of ARIS. This comparison demonstrates the effectiveness of the ontological framework for evaluating modelling tools on how well they can guide selection, implementation and integration of purchased software packages.     

A novel technique to assess chromatin texture using pixel optical densitometry in oral squamous cell carcinoma

Background: The early detection of neoplasia is the prime aim of a diagnostician. Altered chromatin distribution is the earliest microscopic change observed in malignant transformation making it a valuable morphometric parameter. This study was aimed to assess and correlate the progressive changes in chromatin texture from normal to varying grades of premalignancies to malignancy of the oral cavity. Methods: Thirty‐four archival tissue specimens categorized as normal buccal mucosa (5), low‐grade epithelial dysplasia (11), high‐grade epithelial dysplasia (7), and squamous cell carcinoma (11) were stained with Feulgen reaction. Pixel optical densitometry histograms were obtained from analysis of an average of 300 cells/case using ImageJ software. Nine histogram curve characteristics (including area under the curve, area integer percentage, center, centroid) were then analyzed statistically for differences between the four groups. Results: Area integer percentage (P = 0.002), center (P = 0.038), and centroid (P = 0.021) were statistically significant within the four groups The parameters showed a dip in their value from normal to low‐grade dysplasia but showed a steady increase in high‐grade dysplasia and carcinoma. Conclusion: The optical density characteristics had the ability to differentiate the progression of neoplastic changes by extracting information of the chromatin distribution not quantifiable by routine microscopy. The early clumping of chromatin in the periphery in low‐grade dysplasia followed by progressively increasing chromatin aggregates in high‐grade dysplasia and carcinoma correlated well with the shift observed in the histogram center and centroid. The novel pixel optical densitometry technique efficiently predicted malignant transformation. Microsc. Res. Tech. 75:1119–1123, 2012. © 2012 Wiley Periodicals, Inc.     

Motor ability: protective or risk for school injuries?

The study aims were to assess the independent contribution of motor ability to the incidence of school injuries. The study included 2057 pupils in grades 3-6 of primary schools in a city in the north of Israel. A surveillance system gathered information about injuries that occurred on school premises or during school related activities and required medical treatment or caused limitation of usual activities. Children provided information on sensation seeking, self-appraisal of health, academic performance, physical activity, and dominant hand; anthropometric measurements and motor ability tests were performed.The incidence of injury events was 4% (95% CI=3.2-5.0). Injuries increased with increased balance and agility, but there were no differences according to reaction time. No other study variables were associated with the incidence of injuries. Our findings of an increase in the incidence of injuries with better motor ability may express differences in exposure to risk situations between children with better and poorer motor abilities.     

Aspects of the biology of Labeo cylindricus (Pisces: cyprinidae) in Lake Baringo,Kenya

Aspects of the biology of Labeo cylindricus from Lake Baringo were investigated, based on experimental beach seining and gillnetting between August and October 2007. The length–weight relationship indicated the species exhibited positive allometric growth (b = 3.7083), with a condition factor (K) of 0.84 ± 0.0298 S.D. Males dominated fish catches, with an adult sex ratio of males/females of 1:1.7. The fish length‐at‐50% maturity (Lm50) was 127.7 and 126.1 mm total length (TL) for males and females, respectively. Fish age and growth were determined from sagittal otoliths considered to be from the young‐of‐the‐year. Clear circuli observed in the otoliths were used to determine fish age in days. Validation by oxytetracycline marking indicated that a single growth increment (1.1 ± 0.1 SE) is formed daily on fish otoliths. Fish length (TL mm) was correlated significantly with age (in days), TL = 1.398 Age + 26.523 (r² = 0.91). Fish growth was subsequently estimated to be 1.398 mm day^?1. The weights and ages (in days) of fish exhibited a significant (P < 0.05) power relationship defined by the equation: W = 0.0003 Age^2.5804 (r² = 0.88). The results of this study provide parameters that, if appropriately monitored, can be used to predict responses of fish populations in Lake Baringo and elsewhere to human interventions (exploitation) and natural environmental change.