Joaquim Marques de Sá

Applied Statistics Using SPSS, STATISTICA, MATLAB and R

Intended for anyone needing to apply statistical analysis to a large variety of science and engineering problems, this book shows how to use SPSS, MATLAB, STATISTICA and R for data description, statistical inference, classification and regression, factor analysis, survival data and directional statistics. The 2nd edition includes the R language, a new section on bootstrap estimation methods and an improved treatment of tree classifiers, plus additional examples and exercises.

The Porto system for automated cardiotocographic signal analysis

Cardiotocography (CTG) lacks reliability and reproducibility and these problems are believed to be overcome by computer analysis. In this article we describe a system developed for routine clinical automated CTG analysis based on a low cost personal computer. Presently the system has processed 70 ten minute tracings. Fetal heart rate baseline, acceleration--deceleration detection, and long term variability estimation were performed in a satisfactory way.

SEMIAUTOMATED ULTRASONOGRAPHIC MEASUREMENT OF FETAL NUCHAL TRANSLUCENCY USING A COMPUTER SOFTWARE TOOL

Nuchal translucency (NT) thickness measurement has been recently proposed as a part of routine ultrasound scanning during the late first trimester of pregnancy, for the early screening of chromosomal abnormalities. Manual determination of NT is currently performed using electronic calipers placed by the operator in the middle of two echogenic lines displayed on the screen. Therefore, intraobserver and interobserver repeatability can be questioned. This paper presents a software tool that has been developed for achieving this goal in a semiautomatic way, improving the reproducibility of the method.

Improving Deep Neural Network Performance by Reusing Features Trained with Transductive Transference

Transfer Learning is a paradigm in machine learning to solve a target problem by reusing the learning with minor modifications from a different but related source problem. In this paper we propose a novel feature transference approach, especially when the source and the target problems are drawn from different distributions. We use deep neural networks to transfer either low or middle or higher-layer features for a machine trained in either unsupervised or supervised way. Applying this feature transference approach on Convolutional Neural Network and Stacked Denoising Autoencoder on four different datasets, we achieve lower classification error rate with significant reduction in computation time with lower-layer features trained in supervised way and higher-layer features trained in unsupervised way for classifying images of uppercase and lowercase letters dataset.

Modular Neural Network Task Decomposition Via Entropic Clustering

The use of monolithic neural networks (such as a multilayer perceptron) has some drawbacks: e.g. slow learning, weight coupling, the black box effect. These can be alleviated by the use of a modular neural network. The creation of a MNN has three steps: task decomposition, module creation and decision integration. In this paper we propose the use of an entropic clustering algorithm as a way of performing task decomposition. We present experiments on several real world classification problems that show the performance of this approach

Mathematical model for educational simulation of the oxygen delivery to the fetus

Abstract Critical situations in obstetrics and anesthesia of the pregnant woman are rare and associated with a high risk to the woman and fetus involved. Therefore, simulation is a valuable tool in teaching the diagnostic and therapeutic skills in this context. We describe a mathematical model for the oxygen supply to the fetus that was specifically designed for educational simulations. The model reflects the hemodynamics and the oxygen transport. Parameter values for human patients are derived from literature data. We validate the dynamic response of the model, with parameters reflecting the fetal lamb, to various reductions of the uterine and umbilical blood flow.

Using Different Cost Functions to Train Stacked Auto-Encoders

Deep neural networks comprise several hidden layers of units, which can be pre-trained one at a time via an unsupervised greedy approach. A whole network can then be trained (fine-tuned) in a supervised fashion. One possible pre-training strategy is to regard each hidden layer in the network as the input layer of an auto-encoder. Since auto-encoders aim to reconstruct their own input, their training must be based on some cost function capable of measuring reconstruction performance. Similarly, the supervised fine-tuning of a deep network needs to be based on some cost function that reflects prediction performance. In this work we compare different combinations of cost functions in terms of their impact on layer-wise reconstruction performance and on supervised classification performance of deep networks. We employed two classic functions, namely the cross-entropy (CE) cost and the sum of squared errors (SSE), as well as the exponential (EXP) cost, inspired by the error entropy concept. Our results were based on a number of artificial and real-world data sets.

Improving transfer learning accuracy by reusing Stacked Denoising Autoencoders

Transfer learning is a process that allows reusing a learning machine trained on a problem to solve a new problem. Transfer learning studies on shallow architectures show low performance as they are generally based on hand-crafted features obtained from experts. It is therefore interesting to study transference on deep architectures, known to directly extract the features from the input data. A Stacked Denoising Autoencoder (SDA) is a deep model able to represent the hierarchical features needed for solving classification problems. In this paper we study the performance of SDAs trained on one problem and reused to solve a different problem not only with different distribution but also with a different tasks. We propose two different approaches: 1) unsupervised feature transference, and 2) supervised feature transference using deep transfer learning. We show that SDAs using the unsupervised feature transference outperform randomly initialized machines on a new problem. We achieved 7% relative improvement on average error rate and 41% on average computation time to classify typed uppercase letters. In the case of supervised feature transference, we achieved 5.7% relative improvement in the average error rate, by reusing the first and second hidden layer, and 8.5% relative improvement for the average error rate and 54% speed up w.r.t the baseline by reusing all three hidden layers for the same data. We also explore transfer learning between geometrical shapes and canonical shapes, we achieved 7.4% relative improvement on average error rate in case of supervised feature transference approach.

Month of birth and risk of multiple sclerosis in a Portuguese population.

BACKGROUND Several studies analyzing the month of birth (MOB) of multiple sclerosis (MS) patients and the risk of the disease have been published; as a whole, MS patients were found to be predominantly born in spring months, leading to the current assumption that MOB is somewhat related to the risk of MS. OBJECTIVE Estimate the risk of MS by MOB in a Portuguese population. METHODS MS patients sample was obtained from the database of patients attended at our MS clinic and born in the districts of Porto, Braga and Viana do Castelo. The control sample was composed of the live births records in the same time period and geographical area. We applied the Hewitt test for seasonality. RESULTS We found 421 patients that satisfied the conditions to enter the study. The rank-sums for successive 6-month segments indicate the July-December period as of higher incidence; however, the corresponding rank-sum (48) was not statistically significant according to the Hewitt test (p>0.05). CONCLUSION Our data does not support the seasonality hypothesis of MOB as risk factor for MS in Portugal. However we are aware that the analysis of a larger MS sample could shed more light in this issue.

Decision Trees Using the Minimum Entropy-of-Error Principle

Binary decision trees based on univariate splits have traditionally employed so-called impurity functions as a means of searching for the best node splits. Such functions use estimates of the class distributions. In the present paper we introduce a new concept to binary tree design: instead of working with the class distributions of the data we work directly with the distribution of the errors originated by the node splits. Concretely, we search for the best splits using a minimum entropy-of-error (MEE) strategy. This strategy has recently been applied in other areas (e.g. regression, clustering, blind source separation, neural network training) with success. We show that MEE trees are capable of producing good results with often simpler trees, have interesting generalization properties and in the many experiments we have performed they could be used without pruning.