Classification Performance of Multilayer Perceptrons with Different Risk Functionals

In the present paper we assess the performance of information-theoretic inspired risks functionals in multilayer perceptrons with reference to the two most popular ones, Mean Square Error and Cross-Entropy. The information-theoretic inspired risks, recently proposed, are: HS and HR2 are, respectively, the Shannon and quadratic Rényi entropies of the error; ZED is a risk reflecting the error density at zero errors; EXP is a generalized exponential risk, able to mimic a wide variety of risk functionals, including the information-thoeretic ones. The experiments were carried out with multilayer perceptrons on 35 public real-world datasets. All experiments were performed according to the same protocol. The statistical tests applied to the experimental results showed that the ubiquitous mean square error was the less interesting risk functional to be used by multilayer perceptrons. Namely, mean square error never achieved a significantly better classification performance than competing risks. Cross-entropy and EXP were the risks found by several tests to be significantly better than their competitors. Counts of significantly better and worse risks have also shown the usefulness of HS and HR2 for some datasets.

Automatic cymbal classification

Dissertação apresentada na Faculdade de Ciências e Tecnologia da Universidade Nova de Lisboa para a obtenção do grau de Mestre em Engenharia Informática

Partial classification of Lorenz knots: Syllable permutations of torus knots words

We define families of aperiodic words associated to Lorenz knots that arise naturally as syllable permutations of symbolic words corresponding to torus knots. An algorithm to construct symbolic words of satellite Lorenz knots is defined. We prove, subject to the validity of a previous conjecture, that Lorenz knots coded by some of these families of words are hyperbolic, by showing that they are neither satellites nor torus knots and making use of Thurston's theorem. Infinite families of hyperbolic Lorenz knots are generated in this way, to our knowledge, for the first time. The techniques used can be generalized to study other families of Lorenz knots.

Exploiting the bin-class histograms for feature selection on discrete data

In machine learning and pattern recognition tasks, the use of feature discretization techniques may have several advantages. The discretized features may hold enough information for the learning task at hand, while ignoring minor fluctuations that are irrelevant or harmful for that task. The discretized features have more compact representations that may yield both better accuracy and lower training time, as compared to the use of the original features. However, in many cases, mainly with medium and high-dimensional data, the large number of features usually implies that there is some redundancy among them. Thus, we may further apply feature selection (FS) techniques on the discrete data, keeping the most relevant features, while discarding the irrelevant and redundant ones. In this paper, we propose relevance and redundancy criteria for supervised feature selection techniques on discrete data. These criteria are applied to the bin-class histograms of the discrete features. The experimental results, on public benchmark data, show that the proposed criteria can achieve better accuracy than widely used relevance and redundancy criteria, such as mutual information and the Fisher ratio.

Text classification using compression-based dissimilarity measures

Arguably, the most difficult task in text classification is to choose an appropriate set of features that allows machine learning algorithms to provide accurate classification. Most state-of-the-art techniques for this task involve careful feature engineering and a pre-processing stage, which may be too expensive in the emerging context of massive collections of electronic texts. In this paper, we propose efficient methods for text classification based on information-theoretic dissimilarity measures, which are used to define dissimilarity-based representations. These methods dispense with any feature design or engineering, by mapping texts into a feature space using universal dissimilarity measures; in this space, classical classifiers (e.g. nearest neighbor or support vector machines) can then be used. The reported experimental evaluation of the proposed methods, on sentiment polarity analysis and authorship attribution problems, reveals that it approximates, sometimes even outperforms previous state-of-the-art techniques, despite being much simpler, in the sense that they do not require any text pre-processing or feature engineering.

Multi-scale vegetation classification using earth observation data of the Sundarban mangrove forest, Bangladesh.

Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies.

Persistence of Vectobac WDG and Metoprag S-2G against Aedes aegypti larvae using a semi-field bioassay in Rio de Janeiro, Brazil

Persistence of Bacillus thuringiensis var. israelensis (Vectobac WDG) and methoprene (Metoprag S-2G) was evaluated against Aedes aegypti late third instar larvae of the Rockefeller strain in a semi-field bioassay. Tests were performed in Rio de Janeiro, using containers made of plastic, iron, concrete and asbestos, placed in a shaded area. The formulations used were 0.2 g of Vectobac-WDG and 1g of Metoprag S-2G per 100 liters of water in house storage containers. Vectobac WDG was tested twice, in March and in April/May, 2002. In March (temperature ranging from 21.5 to 39.3 ºC), 70-100% mortality was observed by the 7th day and declined abruptly thereafter. No significant differences were observed among the container types. In April/May (18.6 to 34.8 ºC) mortality was higher than 70% to 30-36 days in all cases, except in the iron container (40% mortality on the 12th day). Metoprag S-2G was evaluated in April/May, 2002, and induced mortality higher than 70% up to 15 days in the plastic and iron containers and only seven days in the concrete container. In the asbestos container, maximal mortality was achieved on day one post-treatment (66%). Our results point to a low persistence of both formulations in the weather conditions of Rio de Janeiro.

Semi-Partitioned Scheduling of Fork-Join Tasks using Work-Stealing

13th IEEE/IFIP International Conference on Embedded and Ubiquitous Computing (EUC 2015). 21 to 23, Oct, 2015, Session W1-A: Multiprocessing and Multicore Architectures. Porto, Portugal.

On the Sensor Classification Scheme of Robotic Manipulators

This paper analyzes the signals captured during impacts and vibrations of a mechanical manipulator. The Fourier Transform of eighteen different signals are calculated and approximated by trendlines based on a power law formula. A sensor classification scheme based on the frequency spectrum behavior is presented.

Towards the Combination of Work-Stealing and Semi-Partitioned Scheduling for Parallel Tasks

Paper/Poster presented in Work in Progress Session, 28th GI/ITG International Conference on Architecture of Computing Systems (ARCS 2015). 24 to 26, Mar, 2015. Porto, Portugal.

Towards the Combination of Work-Stealing and Semi-Partitioned Scheduling for Parallel Tasks

Poster presented in Work in Progress Session, 28th GI/ITG International Conference on Architecture of Computing Systems (ARCS 2015). 24 to 26, Mar, 2015. Porto, Portugal.

Towards a classification scheme for musical sounds

Signal Processing: Algorithms, Architectures, Arrangements, and Applications (SPA), 2013

Overhead-aware schedulability evaluation of semi-partitioned real-time schedulers

Presented at IEEE 21st International Conference on Embedded and Real-Time Computing Systems and Applications (RTCSA 2015). 19 to 21, Aug, 2015.

Zero-Time Kidney Histology Predicts Early Graft Function

Our purposes are to determine the impact of histological factors observed in zero-time biopsies on early post transplant kidney allograft function. We specifically want to compare the semi-quantitative Banff Classification of zero time biopsies with quantification of % cortical area fibrosis. Sixty three zero-time deceased donor allograft biopsies were retrospectively semiquantitatively scored using Banff classification. By adding the individual chronic parameters a Banff Chronic Sum (BCS) Score was generated. Percentage of cortical area Picro Sirius Red (%PSR) staining was assessed and calculated with a computer program. A negative linear regression between %PSR/ GFR at 3 year post-transplantation was established (Y=62.08 +-4.6412X; p=0.022). A significant negative correlation between arteriolar hyalinosis (rho=-0.375; p=0.005), chronic interstitial (rho=0.296; p=0.02) , chronic tubular ( rho=0.276; p=0.04) , chronic vascular (rho= -0.360;P=0.007), BCS (rho=-0.413; p=0.002) and GFR at 3 years were found. However, no correlation was found between % PSR, Ci, Ct or BCS. In multivariate linear regression the negative predictive factors of 3 years GFR were: BCS in histological model; donor kidney age, recipient age and black race in clinical model. The BCS seems a good and easy to perform tool, available to every pathologist, with significant predictive short-term value. The %PSR predicts short term kidney function in univariate study and involves extra-routine and expensive-time work. We think that %PSR must be regarded as a research instrument.

Impact of RIFLE Classification in Liver Transplantation

Acute renal failure (ARF) is common after orthotopic liver transplantation (OLT). The aim of this study was to evaluate the prognostic value of RIFLE classification in the development of CKD, hemodialysis requirement, and mortality. Patients were categorized as risk (R), injury (I) or failure (F) according to renal function at day 1, 7 and 21. Final renal function was classified according to K/DIGO guidelines. We studied 708 OLT recipients, transplanted between September 1992 and March 2007; mean age 44 +/- 12.6 yr, mean follow-up 3.6 yr (28.8% > or = 5 yr). Renal dysfunction before OLT was known in 21.6%. According to the RIFLE classification, ARF occurred in 33.2%: 16.8% were R class, 8.5% I class and 7.9% F class. CKD developed in 45.6%, with stages 4 or 5d in 11.3%. Mortality for R, I and F classes were, respectively, 10.9%, 13.3% and 39.3%. Severity of ARF correlated with development of CKD: stage 3 was associated with all classes of ARF, stages 4 and 5d only with severe ARF. Hemodialysis requirement (23%) and mortality were only correlated with the most severe form of ARF (F class). In conclusion, RIFLE classification is a useful tool to stratify the severity of early ARF providing a prognostic indicator for the risk of CKD occurrence and death.