Cross-validation of protein structural class prediction using statistical clustering and neural networks.

We present an approach to predicting protein structural class that uses amino acid composition and hydrophobic pattern frequency information as input to two types of neural networks: (1) a three-layer back-propagation network and (2) a learning vector quantization network. The results of these methods are compared to those obtained from a modified Euclidean statistical clustering algorithm. The protein sequence data used to drive these algorithms consist of the normalized frequency of up to 20 amino acid types and six hydrophobic amino acid patterns. From these frequency values the structural class predictions for each protein (all-alpha, all-beta, or alpha-beta classes) are derived. Examples consisting of 64 previously classified proteins were randomly divided into multiple training (56 proteins) and test (8 proteins) sets. The best performing algorithm on the test sets was the learning vector quantization network using 17 inputs, obtaining a prediction accuracy of 80.2%. The Matthews correlation coefficients are statistically significant for all algorithms and all structural classes. The differences between algorithms are in general not statistically significant. These results show that information exists in protein primary sequences that is easily obtainable and useful for the prediction of protein structural class by neural networks as well as by standard statistical clustering algorithms.     

Using a binaural biomimetic array to identify bottom objects ensonified by echolocating dolphins

The development of a unique dolphin biomimetic sonar produced data that were used to study signal processing methods for object identification. Echoes from four metallic objects proud on the bottom, and a substrate-only condition, were generated by bottlenose dolphins trained to ensonify the targets in very shallow water. Using the two-element ('binaural') receive array, object echo spectra were collected and submitted for identification to four neural network architectures. Identification accuracy was evaluated over two receive array configurations, and five signal processing schemes. The four neural networks included backpropagation, learning vector quantization, genetic learning and probabilistic network architectures. The processing schemes included four methods that capitalized on the binaural data, plus a monaural benchmark process. All the schemes resulted in above-chance identification accuracy when applied to learning vector quantization and backpropagation. Beam-forming or concatenation of spectra from both receive elements outperformed the monaural benchmark, with higher sensitivity and lower bias. Ultimately, best object identification performance was achieved by the learning vector quantization network supplied with beam-formed data. The advantages of multi-element signal processing for object identification are clearly demonstrated in this development of a first-ever dolphin biomimetic sonar.     

Pattern recognition and classification of images of biological macromolecules using artificial neural networks.

The goal of this work was to analyze an image data set and to detect the structural variability within this set. Two algorithms for pattern recognition based on neural networks are presented, one that performs an unsupervised classification (the self-organizing map) and the other a supervised classification (the learning vector quantization). The approach has a direct impact in current strategies for structural determination from electron microscopic images of biological macromolecules. In this work we performed a classification of both aligned but heterogeneous image data sets as well as basically homogeneous but otherwise rotationally misaligned image populations, in the latter case completely avoiding the typical reference dependency of correlation-based alignment methods. A number of examples on chaperonins are presented. The approach is computationally fast and robust with respect to noise. Programs are available through ftp.     

Exploration research on the fusion of multimodal spectrum technology to improve performance of rapid diagnosis scheme for thyroid dysfunction

The spectral fusion by Raman spectroscopy and Fourier infrared spectroscopy combined with pattern recognition algorithms is utilized to diagnose thyroid dysfunction serum, and finds the spectral segment with the highest sensitivity to further advance diagnosis speed. Compared with the single infrared spectroscopy or Raman spectroscopy, the proposal can improve the detection accuracy, and can obtain more spectral features, indicating greater differences between thyroid dysfunction and normal serum samples. For discriminating different samples, principal component analysis (PCA) was first used for feature extraction to reduce the dimension of high‐dimension spectral data and spectral fusion. Then, support vector machine (SVM), back propagation neural network, extreme learning machine and learning vector quantization algorithms were employed to establish the discriminant diagnostic models. The accuracy of spectral fusion of the best analytical model PCA‐SVM, single Raman spectral accuracy and single infrared spectral accuracy is 83.48%, 78.26% and 80%, respectively. The accuracy of spectral fusion is higher than the accuracy of single spectrum in five classifiers. And the diagnostic accuracy of spectral fusion in the range of 2000 to 2500 cm^?1 is 81.74%, which greatly improves the sample measure speed and data analysis speed than analysis of full spectra. The results from our study demonstrate that the serum spectral fusion technique combined with multivariate statistical methods have great potential for the screening of thyroid dysfunction.     

Frequency component selection for an ECoG-based brain-computer interface.

The aim of the present study was to investigate the most significant frequency components in electrocorticogram (ECoG) recordings in order to operate a brain computer interface (BCI). For this purpose the time-frequency ERD/ERS map and the distinction sensitive learning vector quantization (DSLVQ) are applied to ECoG from three subjects, recorded during a self-paced finger movement. The results show that the ERD/ERS pattern found in ECoG generally matches the ERD/ERS pattern found in EEG recordings, but has an increased prevalence of frequency components in the beta range.     

Why neural networks should not be used for HIV-1 protease cleavage site prediction

SUMMARY: Several papers have been published where nonlinear machine learning algorithms, e.g. artificial neural networks, support vector machines and decision trees, have been used to model the specificity of the HIV-1 protease and extract specificity rules. We show that the dataset used in these studies is linearly separable and that it is a misuse of nonlinear classifiers to apply them to this problem. The best solution on this dataset is achieved using a linear classifier like the simple perceptron or the linear support vector machine, and it is straightforward to extract rules from these linear models. We identify key residues in peptides that are efficiently cleaved by the HIV-1 protease and list the most prominent rules, relating them to experimental results for the HIV-1 protease. MOTIVATION: Understanding HIV-1 protease specificity is important when designing HIV inhibitors and several different machine learning algorithms have been applied to the problem. However, little progress has been made in understanding the specificity because nonlinear and overly complex models have been used. RESULTS: We show that the problem is much easier than what has previously been reported and that linear classifiers like the simple perceptron or linear support vector machines are at least as good predictors as nonlinear algorithms. We also show how sets of specificity rules can be generated from the resulting linear classifiers. AVAILABILITY: The datasets used are available at http://www.hh.se/staff/bioinf/     

Democratic population decisions result in robust policy-gradient learning: a parametric study with GPU simulations

High performance computing on the Graphics Processing Unit (GPU) is an emerging field driven by the promise of high computational power at a low cost. However, GPU programming is a non-trivial task and moreover architectural limitations raise the question of whether investing effort in this direction may be worthwhile. In this work, we use GPU programming to simulate a two-layer network of Integrate-and-Fire neurons with varying degrees of recurrent connectivity and investigate its ability to learn a simplified navigation task using a policy-gradient learning rule stemming from Reinforcement Learning. The purpose of this paper is twofold. First, we want to support the use of GPUs in the field of Computational Neuroscience. Second, using GPU computing power, we investigate the conditions under which the said architecture and learning rule demonstrate best performance. Our work indicates that networks featuring strong Mexican-Hat-shaped recurrent connections in the top layer, where decision making is governed by the formation of a stable activity bump in the neural population (a "non-democratic" mechanism), achieve mediocre learning results at best. In absence of recurrent connections, where all neurons "vote" independently ("democratic") for a decision via population vector readout, the task is generally learned better and more robustly. Our study would have been extremely difficult on a desktop computer without the use of GPU programming. We present the routines developed for this purpose and show that a speed improvement of 5x up to 42x is provided versus optimised Python code. The higher speed is achieved when we exploit the parallelism of the GPU in the search of learning parameters. This suggests that efficient GPU programming can significantly reduce the time needed for simulating networks of spiking neurons, particularly when multiple parameter configurations are investigated.     

A protocol for assessment and integration of vegetation maps,with an application to spatial data sets from south‐eastern Australia

Abstract Maps are important tools in natural resource management. Often, there may be multiple maps that represent the same resource, which have been constructed using very different philosophies and methods, at different scales, for different dates and areas. In such cases, conservation planners and other natural resource managers are faced with a choice of map that will best serve their decision making. However, the best available information for a given purpose is often a combination of data from a number of different source maps. In this paper we present a protocol for assessing and integrating multiple maps of vegetation for a particular area of interest. The protocol commences with a consideration of management or policy context and technical issues to determine the basic specifications for the map. It then defines and assesses a set of measurable attributes, representing the concepts of theme, accuracy, precision and currency, for all candidate maps available for compilation. The resulting ranks for accuracy, precision and currency are used to compute a suitability index, which is used to assemble a composite map from the most suitable candidate maps. The final step in the protocol is to display spatial patterns in thematic consistency, accuracy, precision and currency for the composite map. We demonstrate the application of the protocol by constructing a map that discriminates structurally intact native vegetation from cleared land for the whole of New South Wales, south‐eastern Australia. The source data include 46 maps that cover various parts of the region at various scales and which were made at different dates using different methods. The protocol is an explicit and systematic method to evaluate the strengths and weaknesses of alternative data sets. It implements spatial integration in a way that promotes overall accuracy, precision and currency of map data. It also promotes transparent reporting of map limitations, to help map users accommodate risks of map errors in their decision making, and to inform priorities for future survey and mapping.     

组学时代下机器学习方法在临床决策支持中的应用

随着组学技术的不断发展,对于不同层次和类型的生物数据的获取方法日益成熟。在疾病诊治过程中会产生大量数据,通过机器学习等人工智能方法解析复杂、多维、多尺度的疾病大数据,构建临床决策支持工具,辅助医生寻找快速且有效的疾病诊疗方案是非常必要的。在此过程中,机器学习等人工智能方法的选择显得尤为重要。基于此,本文首先从类型和算法角度对临床决策支持领域中常用的机器学习等方法进行简要综述,分别介绍了支持向量机、逻辑回归、聚类算法、Bagging、随机森林和深度学习,对机器学习等方法在临床决策支持中的应用做了相应总结和分类,并对它们的优势和不足分别进行讨论和阐述,为临床决策支持中机器学习等人工智能方法的选择提供有效参考。     

Rapidly measuring the speed of unconscious learning: amnesics learn quickly and happy people slowly

Background

We introduce a method for quickly determining the rate of implicit learning.

Methodology/Principal Findings

The task involves making a binary prediction for a probabilistic sequence over 10 minutes; from this it is possible to determine the influence of events of a different number of trials in the past on the current decision. This profile directly reflects the learning rate parameter of a large class of learning algorithms including the delta and Rescorla-Wagner rules. To illustrate the use of the method, we compare a person with amnesia with normal controls and we compare people with induced happy and sad moods.

Conclusions/Significance

Learning on the task is likely both associative and implicit. We argue theoretically and demonstrate empirically that both amnesia and also transient negative moods can be associated with an especially large learning rate: People with amnesia can learn quickly and happy people slowly.     

A Novel Approach for Lie Detection Based on F-Score and Extreme Learning Machine

A new machine learning method referred to as F-score_ELM was proposed to classify the lying and truth-telling using the electroencephalogram (EEG) signals from 28 guilty and innocent subjects. Thirty-one features were extracted from the probe responses from these subjects. Then, a recently-developed classifier called extreme learning machine (ELM) was combined with F-score, a simple but effective feature selection method, to jointly optimize the number of the hidden nodes of ELM and the feature subset by a grid-searching training procedure. The method was compared to two classification models combining principal component analysis with back-propagation network and support vector machine classifiers. We thoroughly assessed the performance of these classification models including the training and testing time, sensitivity and specificity from the training and testing sets, as well as network size. The experimental results showed that the number of the hidden nodes can be effectively optimized by the proposed method. Also, F-score_ELM obtained the best classification accuracy and required the shortest training and testing time.     

Neuro-fuzzy decision trees

Fuzzy decision trees are powerful, top-down, hierarchical search methodology to extract human interpretable classification rules. However, they are often criticized to result in poor learning accuracy. In this paper, we propose Neuro-Fuzzy Decision Trees (N-FDTs); a fuzzy decision tree structure with neural like parameter adaptation strategy. In the forward cycle, we construct fuzzy decision trees using any of the standard induction algorithms like fuzzy ID3. In the feedback cycle, parameters of fuzzy decision trees have been adapted using stochastic gradient descent algorithm by traversing back from leaf to root nodes. With this strategy, during the parameter adaptation stage, we keep the hierarchical structure of fuzzy decision trees intact. The proposed approach of applying backpropagation algorithm directly on the structure of fuzzy decision trees improves its learning accuracy without compromising the comprehensibility (interpretability). The proposed methodology has been validated using computational experiments on real-world datasets.     

Best harmony, unified RPCL and automated model selection for unsupervised and supervised learning on Gaussian mixtures, three-layer nets and ME-RBF-SVM models

After introducing the fundamentals of BYY system and harmony learning, which has been developed in past several years as a unified statistical framework for parameter learning, regularization and model selection, we systematically discuss this BYY harmony learning on systems with discrete inner-representations. First, we shown that one special case leads to unsupervised learning on Gaussian mixture. We show how harmony learning not only leads us to the EM algorithm for maximum likelihood (ML) learning and the corresponding extended KMEAN algorithms for Mahalanobis clustering with criteria for selecting the number of Gaussians or clusters, but also provides us two new regularization techniques and a unified scheme that includes the previous rival penalized competitive learning (RPCL) as well as its various variants and extensions that performs model selection automatically during parameter learning. Moreover, as a by-product, we also get a new approach for determining a set of 'supporting vectors' for Parzen window density estimation. Second, we shown that other special cases lead to three typical supervised learning models with several new results. On three layer net, we get (i) a new regularized ML learning, (ii) a new criterion for selecting the number of hidden units, and (iii) a family of EM-like algorithms that combines harmony learning with new techniques of regularization. On the original and alternative models of mixture-of-expert (ME) as well as radial basis function (RBF) nets, we get not only a new type of criteria for selecting the number of experts or basis functions but also a new type of the EM-like algorithms that combines regularization techniques and RPCL learning for parameter learning with either least complexity nature on the original ME model or automated model selection on the alternative ME model and RBF nets. Moreover, all the results for the alternative ME model are also applied to other two popular nonparametric statistical approaches, namely kernel regression and supporting vector machine. Particularly, not only we get an easily implemented approach for determining the smoothing parameter in kernel regression, but also we get an alternative approach for deciding the set of supporting vectors in supporting vector machine.     

Classification of cytological smears of the cervix with neuronal methods]

BACKGROUND: Cytological smears obtained from the cervix are routinely examined under the microscope as part of screening programs for the early detection of cervical cancer. The aim of the present study was to investigate whether a simple feature extraction approach using only standard image processing techniques combined with a neural classifier would lead to acceptable results that might serve as a starting point for the development of a fully automated screening system. MATERIALS AND METHODS: Gray-value images of 106 cervical smears (512 x 512 pixels) divided into two groups--inconspicuous (57) and atypical (49)--by an experienced pathologist on the basis of the original smears were employed to evaluate the method. From these images, 31 features quantifying properties of either the cell nucleus or the cytoplasm were extracted. These features were categorized with three different architectures of a neural classifier: learning vector quantization (LVQ), multilayer perceptron (MLP) and a single perceptron. CONCLUSIONS: The results show a reclassification accuracy of about 91% for all three algorithms. Sensitivity was uniform at approximately 78%, and specificity varied between 75% and 91% in the leave-one-out evaluation. These very good results provide strong encouragement for further studies involving PAP scores and colour images.     

Computer aided diagnostic problem solving: Identification of peripheral nerve disorders

AimThe aim was to design and develop a decision support system with a graphical user interface for the prediction of the case of peripheral nerve disorder and to build a classifier using artificial neural networks that can distinguish between carpal tunnel syndrome, neuropathy and normal peripheral nerve conduction.Materials and methodsThe data used were the Nerve Conduction Study data obtained from Kannur Medical College, India. A recurrent neural network and a two-layer feed forward network trained with scaled conjugate gradient back-propagation algorithm were implemented and results were compared.ResultsBoth the networks provided fast convergence and good performance, accuracy being 98.6% and 97.4% for the recurrent neural network and the feed forward networks respectively, the confusion matrix in each case indicated only a few misclassifications. The developed decision support system also gave accurate results in agreement with the specialist's diagnosis and was also useful in storing and viewing the results.DiscussionsIn the field of medicine, programs are being developed that aids in diagnostic decision making by emulating human intelligence such as logical thinking, decision making, learning, etc. The system developed proves useful in combination with other systems in providing diagnostic and predictive medical opinions. It was not meant to replace the specialist, yet it can be used to assist a general practitioner or specialist in diagnosing and predicting patient's condition.ConclusionsThe study proves that artificial neural networks are indeed of value in combination with other systems in providing diagnostic and predictive medical opinions. But the major drawback of these studies, which makes use of the nerve conduction study data are the inherent shortcomings of the interpretation of the results, which include lack of standardization and absence of population-based reference intervals.     

Development and comparative assessment of Raman spectroscopic classification algorithms for lesion discrimination in stereotactic breast biopsies with microcalcifications

Microcalcifications are an early mammographic sign of breast cancer and a target for stereotactic breast needle biopsy. Here, we develop and compare different approaches for developing Raman classification algorithms to diagnose invasive and in situ breast cancer, fibrocystic change and fibroadenoma that can be associated with microcalcifications. In this study, Raman spectra were acquired from tissue cores obtained from fresh breast biopsies and analyzed using a constituent‐based breast model. Diagnostic algorithms based on the breast model fit coefficients were devised using logistic regression, C4.5 decision tree classification, k‐nearest neighbor (k ‐NN) and support vector machine (SVM) analysis, and subjected to leave‐one‐out cross validation. The best performing algorithm was based on SVM analysis (with radial basis function), which yielded a positive predictive value of 100% and negative predictive value of 96% for cancer diagnosis. Importantly, these results demonstrate that Raman spectroscopy provides adequate diagnostic information for lesion discrimination even in the presence of microcalcifications, which to the best of our knowledge has not been previously reported. (© 2013 WILEY‐VCH Verlag GmbH & Co. KGaA, Weinheim)     

Comparison of two unsupervised learning algorithms

The binary decision element described by the decision rule depending upon weight vector w is a model of neuron examined in this paper. The environment of the element is described by some unknown, stationary distribution p(x). The input signals x[n] of the element appear in each step n independently in accordance with the distribution p(x). During an unsupervised learning process the weight vector w[n] is changed on the base of the input vector x[n]. In the paper there are regarded two self-learning algorithms which are stochastic approximation type. For both algorithms the same rule of past experiences neglecting or the rule of weight decrease has been introduced. The first algorithm differs from the other one by a rule of weight increase. It has been proved that only one of these algorithms always leads to the same decision rule in a given environment p(x).This work was done during stay of Dr. L. Bobrowski at the University of Salerno in the frame of Polish-Italian Agreement on Scientific Cooperation     

一种基于支持向量机技术的癫痫脑电棘尖波识别方法

支持向量机是一种基于统计学习理论的新型学习机。文章提出一种基于支持向量机的癫痫脑电特征提取与识别方法,充分发挥其泛化能力强的特点,在与神经网络方法的比较中,表现出较低的漏检率和较好的鲁棒性,有深入研究的价值和良好的应用前景。     

Estimating a Path through a Map of Decision Making

Studies of the evolution of collective behavior consider the payoffs of individual versus social learning. We have previously proposed that the relative magnitude of social versus individual learning could be compared against the transparency of payoff, also known as the “transparency” of the decision, through a heuristic, two-dimensional map. Moving from west to east, the estimated strength of social influence increases. As the decision maker proceeds from south to north, transparency of choice increases, and it becomes easier to identify the best choice itself and/or the best social role model from whom to learn (depending on position on east–west axis). Here we show how to parameterize the functions that underlie the map, how to estimate these functions, and thus how to describe estimated paths through the map. We develop estimation methods on artificial data sets and discuss real-world applications such as modeling changes in health decisions.