Use Of Probabilistic Neural Network Research Articles

Task Accuracy Enhancement for a Surgical Macro-Micro Manipulator With Probabilistic Neural Networks and Uncertainty Minimization

Accurate robot kinematic modelling is a major component for autonomous robot control to guarantee safety and precision during task execution. In surgical robotics complex robotic structures and actuation mechanisms are generally employed, therefore machine learning techniques can be adopted to build the model of the robot. Probabilistic neural networks are a class of learning approaches that provide information about the uncertainty of the learnt models. In this work we compare two different probabilistic neural networks (Bayesian and Evidential Neural Networks) to model the kinematics of a surgical robotic instrument and propose a control strategy based on Hierarchical Quadratic Programming (HQP) capable of exploiting the model uncertainty to improve the accuracy and safety of the controller. Simulation and real world experiments on different autonomous path tracking tasks show that the model uncertainty highly affects the control performances and prove the effectiveness of the proposed controller in improving task execution. <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Note to Practitioners</i> —The push towards reducing invasiveness and patient’s traumas in surgery has lead to the requirement of miniaturized and highly articulated robots. This however comes at the cost of having systems that are hard to model and control, which is one of the major limitations for autonomy in surgical robotics. Machine learning has become very effective in modelling complex systems and probabilistic approaches additionally allow estimating the confidence of the learnt model. In robotics field where high precision is required to perform an autonomous task, like in minimally invasive surgery, the robot model needs to be very accurate and controllers need to guarantee safety in performing the desired task, while satisfying additional motion constraints imposed by the application scenario. This work proposes the use of probabilistic neural networks to model the complexity of a surgical robotic instrument and a control strategy capable of ensuring safety by maximizing model’s confidence and guaranteeing satisfaction of imposed motion constraints. In this work a macro-micro manipulator setup is employed, consisting of an articulated surgical robotic instrument connected to a serial-link manipulator. The proposed modelling and control approaches can be used in any other field where controllers need to highly rely on the robot model due to limitations in using external sensors and where leveraging information about model’s confidence can be beneficial. Currently, the work focuses only on pure kinematic modelling and control, thus neglecting any possible interaction with the environment. Future work will focus on addressing this limitation in order to ensure proper force control and effective autonomy.

Modelling fatigue uncertainty by means of nonconstant variance neural networks

AbstractThe modelling of fatigue using machine learning (ML) has been gaining traction in the engineering community. Among ML techniques, the use of probabilistic neural networks (PNNs) has recently emerged as a candidate for modelling fatigue applications. In this paper, we use PNNs with nonconstant variance to model fatigue. We present two case studies to demonstrate the developed approach. First, we model the fatigue life of cover‐plated beams under constant amplitude loading, and then we model the relationship between random vibration velocity and equivalent stress in process pipework. The two case studies demonstrate that PNNs with nonconstant variance can model the distribution of the data while also considering the variability of both distribution parameters (mean and standard deviation). This shows the potential of PNNs with nonconstant variance in modelling fatigue applications. All the data and code used in this paper are openly available.

Probabilistic neural networks for fluid flow surrogate modeling and data recovery

We consider the use of probabilistic neural networks for fluid flow {surrogate modeling} and data recovery. This framework is constructed by assuming that the target variables are sampled from a Gaussian distribution conditioned on the inputs. Consequently, the overall formulation sets up a procedure to predict the hyperparameters of this distribution which are then used to compute an objective function given training data. We demonstrate that this framework has the ability to provide for prediction confidence intervals based on the assumption of a probabilistic posterior, given an appropriate model architecture and adequate training data. The applicability of the present framework to cases with noisy measurements and limited observations is also assessed. To demonstrate the capabilities of this framework, we consider canonical regression problems of fluid dynamics from the viewpoint of reduced-order modeling and spatial data recovery for four canonical data sets. The examples considered in this study arise from (1) the shallow water equations, (2) a two-dimensional cylinder flow, (3) the wake of NACA0012 airfoil with a Gurney flap, and (4) the NOAA sea surface temperature data set. The present results indicate that the probabilistic neural network not only produces a machine-learning-based fluid flow {surrogate} model but also systematically quantifies the uncertainty therein to assist with model interpretability.

Using Probabilistic Neural Networks to Predict the Localization of Proteins in Cell Compartments

This paper describes the use of probabilistic neural networks to solve problems of bioinformatics by the example of determining the localization of proteins according to their primary structure. The data used are sets of characteristics of amino acid sequences of proteins, obtained by various software tools aimed at finding specific signal sequences, as well as data on where these proteins are localized in cells of two microorganisms — E. coli and S. cerevisiae. The data source is the UCI Machine Learning Repository (http://archive.ics.uci.edu/ml/datasets). The possibility of using probabilistic neural networks to solve this problem is shown, since the classification accuracy of 57.5% and 85.0% for yeast and for bacterial cells is obtained, respectively. The obtained indicators of the accuracy of the classification of the data used exceed those that, according to the literature, were achieved with the use of other recognition methods. It is noted that a high learning rate and the possibility of modification make probabilistic neural networks a promising tool for analyzing bioinformatics data. In silico

Genome-enabled prediction using probabilistic neural network classifiers.

BackgroundMulti-layer perceptron (MLP) and radial basis function neural networks (RBFNN) have been shown to be effective in genome-enabled prediction. Here, we evaluated and compared the classification performance of an MLP classifier versus that of a probabilistic neural network (PNN), to predict the probability of membership of one individual in a phenotypic class of interest, using genomic and phenotypic data as input variables. We used 16 maize and 17 wheat genomic and phenotypic datasets with different trait-environment combinations (sample sizes ranged from 290 to 300 individuals) with 1.4 k and 55 k SNP chips. Classifiers were tested using continuous traits that were categorized into three classes (upper, middle and lower) based on the empirical distribution of each trait, constructed on the basis of two percentiles (15–85 % and 30–70 %). We focused on the 15 and 30 % percentiles for the upper and lower classes for selecting the best individuals, as commonly done in genomic selection. Wheat datasets were also used with two classes. The criteria for assessing the predictive accuracy of the two classifiers were the area under the receiver operating characteristic curve (AUC) and the area under the precision-recall curve (AUCpr). Parameters of both classifiers were estimated by optimizing the AUC for a specific class of interest.ResultsThe AUC and AUCpr criteria provided enough evidence to conclude that PNN was more accurate than MLP for assigning maize and wheat lines to the correct upper, middle or lower class for the complex traits analyzed. Results for the wheat datasets with continuous traits split into two and three classes showed that the performance of PNN with three classes was higher than with two classes when classifying individuals into the upper and lower (15 or 30 %) categories.ConclusionsThe PNN classifier outperformed the MLP classifier in all 33 (maize and wheat) datasets when using AUC and AUCpr for selecting individuals of a specific class. Use of PNN with Gaussian radial basis functions seems promising in genomic selection for identifying the best individuals. Categorizing continuous traits into three classes generally provided better classification than when using two classes, because classification accuracy improved when classes were balanced.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-016-2553-1) contains supplementary material, which is available to authorized users.

The Use of Probabilistic Neural Network and UV Reflectance Spectroscopy as an Objective Cultured Pearl Quality Grading Method

Pearl quality and value are determined as a combination of different features, with mollusk species, nacre thickness, luster, surface, shape, color and pearl size, being the most important. A pearl grader has to quantify visual observations and to assign a grading level to a pearl. The aim of this work was to reduce subjectivity in the assessment of some aspects of pearl quality by using artificial neural networks to predict pearl quality parameters from UV reflectance spectra. Given the good predictability of our previous model that used multilayer perceptron ANN modeling of UVVisible spectra to predict the grade of pearls, we wanted to simplify and improve the model by reducing the spectral input to UV only and by using classifier neural network modelling. It is hypothesized that as UV light is of higher energy than visible light, it may penetrate further into the surface of the pearl, and hence the corresponding UV diffuse reflectance spectrum may provide more information that can be used to assess pearl quality. The developed models were successful in predicting mollusk pearl growing species, pearl and donor color, luster, and surface complexity. The simplified models have been built resulting in more accurate prediction of selected pearl quality parameters when compared with the previous reported model.

Identification of lubricant contamination by biodiesel using vibration analysis and neural network

PurposeThe purpose of this paper is to provide information on lubricant contamination by biodiesel using vibration and neural network.Design/methodology/approachThe possible contamination of lubricants is verified by analyzing the vibration and neural network of a bench test under determinated conditions.FindingsResults have shown that classical signal analysis methods could not reveal any correlation between the signal and the presence of contamination, or contamination grade. On other hand, the use of probabilistic neural network (PNN) was very successful in the identification and classification of contamination and its grade.Research limitations/implicationsThis study was done for some specific kinds of biodiesel. Other types of biodiesel could be analyzed.Practical implicationsContamination information is presented in the vibration signal, even if it is not evident by classical vibration analysis. In addition, the use of PNN gives a relatively simple and easy‐to‐use detection tool with good confidence. The training process is fast, and allows implementation of an adaptive training algorithm.Originality/valueThis research could be extended to an internal combustion engine in order to verify a possible contamination by biodiesel.

Use of probabilistic neural networks to solve classification problems in surgery

The paper considers the classification of peritonitis-stricken patients with regard to the outcome of the operation and the probable result of the treatment. A probabilistic neural network is offered as a classifier.

Feature extraction of partial discharge signals using the wavelet packet transform and classification with a probabilistic neural network

Partial discharge (PD) classification in power cable accessories and high voltage equipment in general is essential in evaluating the severity of the damage in the insulation. In this article, the PD classification was realised as a two-fold process. Firstly, measurements taken from a high-frequency current transformer (HFCT) sensor were represented as features by means of a transformation to the classifier and secondly, the probabilistic neural network (PNN) classifier itself was capable of effectively recognising features coming from different types of discharges. The feature that was used as a fingerprint for PD characterisation was extracted from the moments of the probability density function (PDF) of the wavelet coefficients at various scales, obtained through the wavelet packets transformation. The PNN classifier was used to classify the PDs and assess the suitability of this feature vector in PD classification. Four types of artificial PDs were created in a high voltage laboratory, namely corona discharge in air, floating discharge in oil, internal discharge in oil and surface discharge in air, at different applied voltages, and were used to train the PNN algorithm. The results obtained here (97.49, 91.9, 100 and 99.8% for the corona, the floating, the internal and the surface discharges, respectively) are very encouraging for the use of PNN in PD classification with this particular feature vector. This article suggests a feature extraction and classification algorithm for PD classification, which when combined together reduced the dimensionality of the feature space to a manageable dimension, and achieved very high levels of classification.

Detection of Resistivity for Antibiotics by Probabilistic Neural Networks

This paper presents the use of probabilistic neural networks (PNNs) for detection of resistivity for antibiotics (resistant and sensitive). The PNN is trained on the resistivity or sensitivity to the antibiotics of each record in the Salmonella database. Estimation of the whole parameter space for the PNN was performed by the maximum-likelihood (ML) estimation method. The expectation-maximization (EM) approach can help to achieve the ML estimation via iterative computation. Resistivity and sensitivity of the three antibiotics (ampicillin, chloramphenicol disks and trimethoprim-sulfamethoxazole) were classified with high accuracies by the PNN. The obtained results demonstrated the success of the PNN to help in detection of resistivity for antibiotics.

Use of probabilistic neural network to construct early warning model for business financial distress

Because Artificial Neural Network possesses powerful capabilities in filtering and learning adaptability and functions of multi-I/O, thus, this method is expected to be more efficient than those of classical statistics modeling constructs. In the domain of Artificial Neural Network, Probabilistic Neural Network demonstrates broader and much more generalized capabilities, as of now, the latter has been successfully applied to different fields. In light of this, this thesis adopts Probabilistic Neural Network to proceed constructs for Early Warning Model of Business Financial Distress. And this thesis started by collecting relevant information from selected 50 financial crisis prone companies and those respective pairing companies in Taiwan, then followed with Principal Component Analysis in order to reduce model input variables. Finally, through constructing respective forecasting models with Probabilistic Neural Network, Back-Propagation Networks and Logistic Regression, researcher can compare the three models as intended. Results showed that, model forecasting on Early Warning Model of Business Financial Distress by Probabilistic Neural Network with its parameters fine-tuned, exhibited excellent performance in this regard.

Novelty detection with constructive probabilistic neural networks

This paper investigates the use of probabilistic neural networks trained with the dynamic decay adjustment algorithm (PNN–DDA) for novelty detection tasks. PNN–DDA is a fast, constructive neural model originally developed and investigated for standard classification tasks. The training algorithm is controlled by two parameters, θ + and θ - . Simulations employing four data sets from the UCI machine learning repository are reported. The results show that parameter θ - considerably influences the performance of PNN–DDA for novelty detection, and furthermore, that PNN–DDA achieves performance comparable to NNDD with the advantage of producing much smaller classifiers.

Approach into the use of probabilistic neural networks for automated classification of tool malfunctions in broaching

The condition of broaching tools has crucial importance for the surface quality of the machined components. If undetected, tool malfunctions such as wear, chipping and breakage of cutting teeth can result in severe damage or even scrapping expensive components, with direct implications on increasing the overall manufacturing costs. In contrast with other machining operations, broaching is characterised by non-symmetric distributions of cutting forces vs. time, making more difficult the task of recognising tool malfunctions. The paper reports on a methodology to automatically detect and classify tool malfunctions in broaching. The method was demonstrated through the use of time domain distribution of the push-off cutting force as a key sensory signal to monitor broaching tool condition when machining a nickel-based aerospace alloy. The characteristic features of the sensory signals have been extracted using in-house-developed programs and, afterwards, used to train and test a probabilistic neural network that enables automated classification of tools with fresh, worn, chipped and broken teeth. Inputting new pattern characteristics to the main categories of tool malfunctions, the system successfully classified them even when variations of signal amplitude and ranking of malfunctioned teeth occurred.

Using probabilistic neural networks to model the toxicity of chemicals to the fathead minnow ( Pimephales promelas): A study based on 865 compounds

We investigate the use of probabilistic neural networks (PNN) to model the acute toxicity (96-hr LC50) to the fathead minnow ( Pimephales promelas) based on a 865 chemicals data set. In contrast to most other toxicological models, the octanol/water partition coefficient is not used as input parameter. The information fed into the neural network is solely based on simple molecular descriptors as can be derived from the chemicals' structures and indicates the potential of this approach as general methodology for the estimation of toxicological effects of chemicals.