Search

In this paper, the implementation of artificial neural networks (multilayer perceptron [MLP] and radial base functions [RBF]) and the upgraded Markov chain model have been studied and performed to identify the human behavior patterns during rock, paper, and scissors game. The main motivation of this research is the design and construction of an intelligent robot with the ability to defeat a human opponent. MATLAB software has been used to implement intelligent algorithms. After implementing the algorithms, their effectiveness in detecting human behavior pattern has been investigated. To ensure the ideal performance of the implemented model, each player played with the desired algorithms in three different stages. The results showed that the percentage of winning computer with MLP and RBF neural networks and upgraded Markov model, on average in men and women is 59%, 76.66%, and 75%, respectively. Obtained results clearly indicate a very good performance of the RBF neural network and the upgraded Markov model in the mental modeling of the human opponent in the game of rock, paper, and scissors. In the end, the designed game has been employed in both hardware and software which include the Zana intelligent robot and a digital version with a graphical user interface design on the stand. To the best knowledge of the authors, the precision of novel presented method for determining human behavior patterns was the highest precision among all of the previous studies.

We demonstrate the use of a convolutional neural network to perform neutron-gamma pulse shape discrimination, where the only inputs to the network are the raw digitised SiPM signals from a dual scintillator detector element made of 6LiF:ZnS(Ag) scintillator and PVT plastic. A realistic labelled dataset was created to train the network by exposing the detector to an AmBe source, and a data-driven method utilsing a separate PMT was used to assign labels to the recorded signals. This approach is compared to the charge integration and continuous wavelet transform methods and is found to provide superior levels of discrimination, achieving an AUC of 0.995 +/- 0.003. We find that the neural network is capable of extracting interpretable features directly from the raw data. In addition, by visualising the high-dimensional representations of the network with the t-SNE algorithm, we discover that not only is this method robust to minor mislabeling of the training dataset but that it is possible to identify an underlying substructure within the signals that goes beyond the original labelling. This technique could be utilised to explore and cluster complex, raw detector data in a novel way that may reveal more insights than standard analysis methods., 13 pages, 8 figures, submitted to Nucl. Instr. Meth. A

Funder: Royal Society; doi: http://dx.doi.org/10.13039/501100000288, Funder: Trinity College, University of Cambridge; doi: http://dx.doi.org/10.13039/501100000727, Funder: John Fell Fund, University of Oxford; doi: http://dx.doi.org/10.13039/501100004789, We use a deep neural network (NN) to detect and place region-of-interest (ROI) boxes around ultracold atom clouds in absorption and fluorescence images—with the ability to identify and bound multiple clouds within a single image. The NN also outputs segmentation masks that identify the size, shape and orientation of each cloud from which we extract the clouds’ Gaussian parameters. This allows 2D Gaussian fits to be reliably seeded thereby enabling fully automatic image processing. The method developed performs significantly better than a more conventional method based on a standardized image analysis library (Scikit-image) both for identifying ROI and extracting Gaussian parameters.

Significance: Early detection of oral cancer is vital for high-risk patients, and machine learning-based automatic classification is ideal for disease screening. However, current datasets collected from high-risk populations are unbalanced and often have detrimental effects on the performance of classification. Aim: To reduce the class bias caused by data imbalance. Approach: We collected 3851 polarized white light cheek mucosa images using our customized oral cancer screening device. We use weight balancing, data augmentation, undersampling, focal loss, and ensemble methods to improve the neural network performance of oral cancer image classification with the imbalanced multi-class datasets captured from high-risk populations during oral cancer screening in low-resource settings. Results: By applying both data-level and algorithm-level approaches to the deep learning training process, the performance of the minority classes, which were difficult to distinguish at the beginning, has been improved. The accuracy of “premalignancy” class is also increased, which is ideal for screening applications. Conclusions: Experimental results show that the class bias induced by imbalanced oral cancer image datasets could be reduced using both data- and algorithm-level methods. Our study may provide an important basis for helping understand the influence of unbalanced datasets on oral cancer deep learning classifiers and how to mitigate.

Significance: Sub-diffuse optical properties may serve as useful cancer biomarkers, and wide-field heatmaps of these properties could aid physicians in identifying cancerous tissue. Sub-diffuse spatial frequency domain imaging (sd-SFDI) can reveal such wide-field maps, but the current time cost of experimentally validated methods for rendering these heatmaps precludes this technology from potential real-time applications. Aim: Our study renders heatmaps of sub-diffuse optical properties from experimental sd-SFDI images in real time and reports these properties for cancerous and normal skin tissue subtypes. Approach: A phase function sampling method was used to simulate sd-SFDI spectra over a wide range of optical properties. A machine learning model trained on these simulations and tested on tissue phantoms was used to render sub-diffuse optical property heatmaps from sd-SFDI images of cancerous and normal skin tissue. Results: The model accurately rendered heatmaps from experimental sd-SFDI images in real time. In addition, heatmaps of a small number of tissue samples are presented to inform hypotheses on sub-diffuse optical property differences across skin tissue subtypes. Conclusion: These results bring the overall process of sd-SFDI a fundamental step closer to real-time speeds and set a foundation for future real-time medical applications of sd-SFDI such as image guided surgery.

Significance: In general, image reconstruction methods used in diffuse optical tomography (DOT) are based on diffusion approximation, and they consider the breast tissue as a homogenous, semi-infinite medium. However, the semi-infinite medium assumption used in DOT reconstruction is not valid when the chest wall is underneath the breast tissue. Aim: We aim to reduce the chest wall’s effect on the estimated average optical properties of breast tissue and obtain accurate forward model for DOT reconstruction. Approach: We propose a deep learning-based neural network approach where a convolution neural network (CNN) is trained to simultaneously obtain accurate optical property values for both the breast tissue and the chest wall. Results: The CNN model shows great promise in reducing errors in estimating the optical properties of the breast tissue in the presence of a shallow chest wall. For patient data, the CNN model predicted the breast tissue optical absorption coefficient, which was independent of chest wall depth. Conclusions: Our proposed method can be readily used in DOT and diffuse spectroscopy measurements to improve the accuracy of estimated tissue optical properties.

Significance: Fourier ptychography (FP) is a computational imaging approach that achieves high-resolution reconstruction. Inspired by neural networks, many deep-learning-based methods are proposed to solve FP problems. However, the performance of FP still suffers from optical aberration, which needs to be considered. Aim: We present a neural network model for FP reconstructions that can make proper estimation toward aberration and achieve artifact-free reconstruction. Approach: Inspired by the iterative reconstruction of FP, we design a neural network model that mimics the forward imaging process of FP via TensorFlow. The sample and aberration are considered as learnable weights and optimized through back-propagation. Especially, we employ the Zernike terms instead of aberration to decrease the optimization freedom of pupil recovery and perform a high-accuracy estimation. Owing to the auto-differentiation capabilities of the neural network, we additionally utilize total variation regularization to improve the visual quality. Results: We validate the performance of the reported method via both simulation and experiment. Our method exhibits higher robustness against sophisticated optical aberrations and achieves better image quality by reducing artifacts. Conclusions: The forward neural network model can jointly recover the high-resolution sample and optical aberration in iterative FP reconstruction. We hope our method that can provide a neural-network perspective to solve iterative-based coherent or incoherent imaging problems.

Reliable methods for the classification and quantification of quantum entanglement are fundamental to understanding its exploitation in quantum technologies. One such method, known as Separable Neural Network Quantum States (SNNS), employs a neural network inspired parameterisation of quantum states whose entanglement properties are explicitly programmable. Combined with generative machine learning methods, this ansatz allows for the study of very specific forms of entanglement which can be used to infer/measure entanglement properties of target quantum states. In this work, we extend the use of SNNS to mixed, multipartite states, providing a versatile and efficient tool for the investigation of intricately entangled quantum systems. We illustrate the effectiveness of our method through a number of examples, such as the computation of novel tripartite entanglement measures, and the approximation of ultimate upper bounds for qudit channel capacities., Comment: 14 pages, 7 figures

Automatic segmentation of bones in computed tomography (CT) images is used for instance in beam hardening correction algorithms where it improves the accuracy of resulting CT numbers. Of special interest are pelvic bones, which—because of their strong attenuation—affect the accuracy of brachytherapy in this region. This work evaluated the performance of the JJ2016 algorithm with the performance of MK2014v2 and JS2018 algorithms; all these algorithms were developed by authors. Visual comparison, and, in the latter case, also Dice similarity coefficients derived from the ground truth were used. It was found that the 3D-based JJ2016 performed better than the 2D-based MK2014v2, mainly because of the more accurate hole filling that benefitted from information in adjacent slices. The neural network-based JS2018 outperformed both traditional algorithms. It was, however, limited to the resolution of 1283 owing to the limited amount of memory in the graphical processing unit (GPU). Funding: VetenskapsradetSwedish Research Council [VR-NT 2016-05033]

Funder: Cantab Capital Institute for the Mathematics of Information, Funder: Alan Turing Institute; doi: https://doi.org/10.13039/100012338, In recent years the use of convolutional layers to encode an inductive bias (translational equivariance) in neural networks has proven to be a very fruitful idea. The successes of this approach have motivated a line of research into incorporating other symmetries into deep learning methods, in the form of group equivariant convolutional neural networks. Much of this work has been focused on roto-translational symmetry of R d , but other examples are the scaling symmetry of R d and rotational symmetry of the sphere. In this work, we demonstrate that group equivariant convolutional operations can naturally be incorporated into learned reconstruction methods for inverse problems that are motivated by the variational regularisation approach. Indeed, if the regularisation functional is invariant under a group symmetry, the corresponding proximal operator will satisfy an equivariance property with respect to the same group symmetry. As a result of this observation, we design learned iterative methods in which the proximal operators are modelled as group equivariant convolutional neural networks. We use roto-translationally equivariant operations in the proposed methodology and apply it to the problems of low-dose computerised tomography reconstruction and subsampled magnetic resonance imaging reconstruction. The proposed methodology is demonstrated to improve the reconstruction quality of a learned reconstruction method with a little extra computational cost at training time but without any extra cost at test time.

Purpose: Deep learning (DL) algorithms have shown promising results for brain tumor segmentation in MRI. However, validation is required prior to routine clinical use. We report the first randomized and blinded comparison of DL and trained technician segmentations. Approach: We compiled a multi-institutional database of 741 pretreatment MRI exams. Each contained a postcontrast T1-weighted exam, a T2-weighted fluid-attenuated inversion recovery exam, and at least one technician-derived tumor segmentation. The database included 729 unique patients (470 males and 259 females). Of these exams, 641 were used for training the DL system, and 100 were reserved for testing. We developed a platform to enable qualitative, blinded, controlled assessment of lesion segmentations made by technicians and the DL method. On this platform, 20 neuroradiologists performed 400 side-by-side comparisons of segmentations on 100 test cases. They scored each segmentation between 0 (poor) and 10 (perfect). Agreement between segmentations from technicians and the DL method was also evaluated quantitatively using the Dice coefficient, which produces values between 0 (no overlap) and 1 (perfect overlap). Results: The neuroradiologists gave technician and DL segmentations mean scores of 6.97 and 7.31, respectively (p

Significance: The non-destructive characterization of cardiac tissue composition provides essential information for both planning and evaluating the effectiveness of surgical interventions such as ablative procedures. Although several methods of tissue characterization, such as optical coherence tomography and fiber-optic confocal microscopy, show promise, many barriers exist that reduce effectiveness or prevent adoption, such as time delays in analysis, prohibitive costs, and limited scope of application. Developing a rapid, low-cost non-destructive means of characterizing cardiac tissue could improve planning, implementation, and evaluation of cardiac surgical procedures. Aim: To determine whether a new light-scattering spectroscopy (LSS) system that analyzes spectra via neural networks is capable of predicting the nuclear densities (NDs) of ventricular tissues. Approach: We developed an LSS system with a fiber-optics probe and applied it for measurements on cardiac tissues from an ovine model. We quantified the ND in the cardiac tissues using fluorescent labeling, confocal microscopy, and image processing. Spectra acquired from the same cardiac tissues were analyzed with spectral clustering and convolutional neural networks (CNNs) to assess the feasibility of characterizing the ND of tissue via LSS. Results: Spectral clustering revealed distinct groups of spectra correlated to ranges of ND. CNNs classified three groups of spectra with low, medium, or high ND with an accuracy of 95.00±11.77% (mean and standard deviation). Our analyses revealed the sensitivity of the classification accuracy to wavelength range and subsampling of spectra. Conclusions: LSS and machine learning are capable of assessing ND in cardiac tissues. We suggest that the approach is useful for the diagnosis of cardiac diseases associated with changes of ND, such as hypertrophy and fibrosis.

Biomedical photoacoustic tomography, which can provide high resolution 3D soft tissue images based on the optical absorption, has advanced to the stage at which translation from the laboratory to clinical settings is becoming possible. The need for rapid image formation and the practical restrictions on data acquisition that arise from the constraints of a clinical workflow are presenting new image reconstruction challenges. There are many classical approaches to image reconstruction, but ameliorating the effects of incomplete or imperfect data through the incorporation of accurate priors is challenging and leads to slow algorithms. Recently, the application of Deep Learning, or deep neural networks, to this problem has received a great deal of attention. This paper reviews the literature on learned image reconstruction, summarising the current trends, and explains how these new approaches fit within, and to some extent have arisen from, a framework that encompasses classical reconstruction methods. In particular, it shows how these new techniques can be understood from a Bayesian perspective, providing useful insights. The paper also provides a concise tutorial demonstration of three prototypical approaches to learned image reconstruction. The code and data sets for these demonstrations are available to researchers. It is anticipated that it is in in vivo applications - where data may be sparse, fast imaging critical and priors difficult to construct by hand - that Deep Learning will have the most impact. With this in mind, the paper concludes with some indications of possible future research directions.

Quasi-static ultrasound elastography (USE) is an imaging modality that measures deformation (i.e. strain) of tissue in response to an applied mechanical force. In USE, the strain modulus is traditionally obtained by deriving the displacement field estimated between a pair of radio-frequency data. In this work we propose a recurrent network architecture with convolutional long-short-term memory decoder blocks to improve displacement estimation and spatio-temporal continuity between time series ultrasound frames. The network is trained in an unsupervised way, by optimising a similarity metric between the reference and compressed image. Our training loss is also composed of a regularisation term that preserves displacement continuity by directly optimising the strain smoothness, and a temporal continuity term that enforces consistency between successive strain predictions. In addition, we propose an open-access in vivo database for quasi-static USE, which consists of radio-frequency data sequences captured on the arm of a human volunteer. Our results from numerical simulation and in vivo data suggest that our recurrent neural network can account for larger deformations, as compared with two other feed-forward neural networks. In all experiments, our recurrent network outperformed the state-of-the-art for both learning-based and optimisation-based methods, in terms of elastographic signal-to-noise ratio, strain consistency, and image similarity. Finally, our open-source code provides a 3D-slicer visualisation module that can be used to process ultrasound RF frames in real-time, at a rate of up to 20 frames per second, using a standard GPU.

Non-contact vital sign monitoring enables the estimation of vital signs, such as heart rate, respiratory rate and oxygen saturation (SpO2), by measuring subtle color changes on the skin surface using a video camera. For patients in a hospital ward, the main challenges in the development of continuous and robust non-contact monitoring techniques are the identification of time periods and the segmentation of skin regions of interest (ROIs) from which vital signs can be estimated. We propose a deep learning framework to tackle these challenges. Approach: This paper presents two convolutional neural network (CNN) models. The first network was designed for detecting the presence of a patient and segmenting the patient’s skin area. The second network combined the output from the first network with optical flow for identifying time periods of clinical intervention so that these periods can be excluded from the estimation of vital signs. Both networks were trained using video recordings from a clinical study involving 15 pre-term infants conducted in the high dependency area of the neonatal intensive care unit (NICU) of the John Radcliffe Hospital in Oxford, UK. Main results: Our proposed methods achieved an accuracy of 98.8% for patient detection, a mean intersection-over-union (IOU) score of 88.6% for skin segmentation and an accuracy of 94.5% for clinical intervention detection using two-fold cross validation. Our deep learning models produced accurate results and were robust to different skin tones, changes in light conditions, pose variations and different clinical interventions by medical staff and family visitors. Significance: Our approach allows cardio-respiratory signals to be continuously derived from the patient’s skin during which the patient is present and no clinical intervention is undertaken.

In the context of epilepsy monitoring, electroencephalography (EEG) remains the modality of choice. Functional near-infrared spectroscopy (fNIRS) is a relatively innovative modality that cannot only characterize hemodynamic profiles of seizures but also allow for long-term recordings. We employ deep learning methods to investigate the benefits of integrating fNIRS measures for seizure detection. We designed a deep recurrent neural network with long short-term memory units and subsequently validated it using the CHBMIT scalp EEG database—a compendium of 896 h of surface EEG seizure recordings. After validating our network using EEG, fNIRS, and multimodal data comprising a corpus of 89 seizures from 40 refractory epileptic patients was used as model input to evaluate the integration of fNIRS measures. Following heuristic hyperparameter optimization, multimodal EEG-fNIRS data provide superior performance metrics (sensitivity and specificity of 89.7% and 95.5%, respectively) in a seizure detection task, with low generalization errors and loss. False detection rates are generally low, with 11.8% and 5.6% for EEG and multimodal data, respectively. Employing multimodal neuroimaging, particularly EEG-fNIRS, in epileptic patients, can enhance seizure detection performance. Furthermore, the neural network model proposed and characterized herein offers a promising framework for future multimodal investigations in seizure detection and prediction.

This paper presents the development and evaluation of three fuzzy neural network (FNN) models for a full-scale anaerobic digestion system treating paper-mill wastewater. The aim was the investigation of feasibility of the approach-based control system for the prediction of effluent quality and biogas production from an internal circulation (IC) anaerobic reactor system. To improve FNN performance, fuzzy subtractive clustering was used to identify model's architecture and optimize fuzzy rule, and a total of 5 rules were extracted in the IF-THEN format. Findings of this study clearly indicated that, compared to NN models, FNN models had smaller RMSE and MAPE as well as bigger R for the testing datasets than NN models. The proposed FNN model produced smaller deviations and exhibited a superior predictive performance on forecasting of both effluent quality and biogas (methane) production rates with satisfactory determination coefficients greater than 0.90. From the results, it was concluded that FNN modeling could be applied in IC anaerobic reactor for predicting the biodegradation and biogas production using paper-mill wastewater.

Significance: The use of optofluidic time-stretch flow cytometry enables extreme-throughput cell imaging but suffers from the difficulties of capturing and processing a large amount of data. As significant amounts of continuous image data are generated, the images require identification with high speed. Aim: We present an intelligent cell phenotyping framework for high-throughput optofluidic time-stretch microscopy based on the XGBoost algorithm, which is able to classify obtained cell images rapidly and accurately. The applied image recognition consists of density-based spatial clustering of applications with noise outlier detection, histograms of oriented gradients combining gray histogram fused feature, and XGBoost classification. Approach: We tested the ability of this framework against other previously proposed or commonly used algorithms to phenotype two groups of cell images. We quantified their performances with measures of classification ability and computational complexity based on AUC and test runtime. The tested cell image datasets were acquired from high-throughput imaging of over 20,000 drug-treated and untreated cells with an optofluidic time-stretch microscope. Results: The framework we built beats other methods with an accuracy of over 97% and a classification frequency of 3000 cells/s. In addition, we determined the optimal structure of training sets according to model performances under different training set components. Conclusions: The proposed XGBoost-based framework acts as a promising solution to processing large flow image data. This work provides a foundation for future cell sorting and clinical practice of high-throughput imaging cytometers.

Pointing is one of the communicative actions that infants acquire during their first year of life. Based on a hypothesis that early pointing is triggered by emergent reaching behavior toward objects placed at out‐of‐reach distances, we proposed a neural network model that acquires reaching without explicit representation of ‘targets’. The proposed model controls a two‐joint arm in a horizontal plane, and it learns a loop of internal forward and inverse transformations; the former predicts the visual feedback of hand position and the latter generates motor commands from the visual input through random generation of the motor commands. In the proposed model, the motor output and visual input were represented by broadly tuned neural units. Even though explicit ‘targets’ were not presented during learning, the simulation successfully generated reaching toward visually presented objects at within‐reach and out‐of‐reach distances.

In the course of infecting their hosts, pathogenic bacteria secrete numerous effectors, namely, bacterial proteins that pervert host cell biology. Many Gram-negative bacteria, including context-dependent human pathogens, use a type IV secretion system (T4SS) to translocate effectors directly into the cytosol of host cells. Various type IV secreted effectors (T4SEs) have been experimentally validated to play crucial roles in virulence by manipulating host cell gene expression and other processes. Consequently, the identification of novel effector proteins is an important step in increasing our understanding of host-pathogen interactions and bacterial pathogenesis. Here, we train and compare six machine learning models, namely, Naive Bayes (NB), K-nearest neighbor (KNN), logistic regression (LR), random forest (RF), support vector machines (SVMs) and multilayer perceptron (MLP), for the identification of T4SEs using 10 types of selected features and 5-fold cross-validation. Our study shows that: (1) including different but complementary features generally enhance the predictive performance of T4SEs; (2) ensemble models, obtained by integrating individual single-feature models, exhibit a significantly improved predictive performance and (3) the 'majority voting strategy' led to a more stable and accurate classification performance when applied to predicting an ensemble learning model with distinct single features. We further developed a new method to effectively predict T4SEs, Bastion4 (Bacterial secretion effector predictor for T4SS), and we show our ensemble classifier clearly outperforms two recent prediction tools. In summary, we developed a state-of-the-art T4SE predictor by conducting a comprehensive performance evaluation of different machine learning algorithms along with a detailed analysis of single- and multi-feature selections.

We develop neural-network-based methods for classifying plaque types in clinical intravascular optical coherence tomography (IVOCT) images of coronary arteries. A single IVOCT pullback can consist of >500 microscopic-resolution images, creating both a challenge for physician interpretation during an interventional procedure and an opportunity for automated analysis. In the proposed method, we classify each A-line, a datum element that better captures physics and pathophysiology than a voxel, as a fibrous layer followed by calcification (fibrocalcific), a fibrous layer followed by a lipidous deposit (fibrolipidic), or other. For A-line classification, the usefulness of a convolutional neural network (CNN) is compared with that of a fully connected artificial neural network (ANN). A total of 4469 image frames across 48 pullbacks that are manually labeled using consensus labeling from two experts are used for training, evaluation, and testing. A 10-fold cross-validation using held-out pullbacks is applied to assess classifier performance. Noisy A-line classifications are cleaned by applying a conditional random field (CRF) and morphological processing to pullbacks in the en-face view. With CNN (ANN) approaches, we achieve an accuracy of 77.7 % ± 4.1 % (79.4 % ± 2.9 % ) for fibrocalcific, 86.5 % ± 2.3 % (83.4 % ± 2.6 % ) for fibrolipidic, and 85.3 % ± 2.5 % (82.4 % ± 2.2 % ) for other, across all folds following CRF noise cleaning. The results without CRF cleaning are typically reduced by 10% to 15%. The enhanced performance of the CNN was likely due to spatial invariance of the convolution operation over the input A-line. The predicted en-face classification maps of entire pullbacks agree favorably to the annotated counterparts. In some instances, small error regions are actually hard to call when re-examined by human experts. Even in worst-case pullbacks, it can be argued that the results will not negatively impact usage by physicians, as there is a preponderance of correct calls.

In the organosolv pulping of the oil palm fronds, the influence of the operational variables of the pulping reactor (viz. cooking temperature and time, ethanol and NaOH concentration) on the properties of the resulting pulp (yield and kappa number) and paper sheets (tensile index and tear index) was investigated using a wavelet neural network model. The experimental results with error less than 0.0965 (in terms of MSE) were produced, and were then compared with those obtained from the response surface methodology. Performance assessment indicated that the neural network model possessed superior predictive ability than the polynomial model, since a very close agreement between the experimental and the predicted values was obtained.

The segmentation of organs at risk is a crucial and time-consuming step in radiotherapy planning. Good automatic methods can significantly reduce the time clinicians have to spend on this task. Due to its variability in shape and low contrast to surrounding structures, segmenting the parotid gland is challenging. Motivated by the recent success of deep learning, we study the use of two-dimensional (2-D), 2-D ensemble, and three-dimensional (3-D) U-Nets for segmentation. The mean Dice similarity to ground truth is ∼0.83 for all three models. A patch-based approach for class balancing seems promising for false-positive reduction. The 2-D ensemble and 3-D U-Net are applied to the test data of the 2015 MICCAI challenge on head and neck autosegmentation. Both deep learning methods generalize well onto independent data (Dice 0.865 and 0.88) and are superior to a selection of model- and atlas-based methods with respect to the Dice coefficient. Since appropriate reference annotations are essential for training but often difficult and expensive to obtain, it is important to know how many samples are needed for training. We evaluate the performance after training with different-sized training sets and observe no significant increase in the Dice coefficient for more than 250 training cases.

This paper addresses the issues of neural network model development and maintenance in the context of a complex task taken from the papermaking industry. In particular, it describes a comparison study of early stopping techniques and model selection, both to optimise neural network models for generalisation performance. The results presented here show that early stopping via use of a Bayesian model evidence measure is a viable way of optimising performance while also making maximum use of all the data. In addition, they show that ten-fold cross-validation performs well as a model selector and as an estimator of prediction accuracy. These results are important in that they show how neural network models may be optimally trained and selected for highly complex industrial tasks where the data are noisy and limited in number.

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior The dimensional stability of the paper may change due to exchange moisture middle, releasing the stress acquired in the manufacturing process. One result of this release of tension is the diagonal curl, or also known as twist. In the paper industry, laboratory measurements are important to ensure the final quality of the material delivered to the customer. Particularly in twist measurement, sampling are made at the end of each jumbo roll, demanding time for any corrections, by operation. With these assumptions, this study aims to do a sensitivity analysis of the different input variables of an industrial board machine, along with some laboratory measurements, related with the property of the paper called diagonal curl (twist) and propose artificial neural networks for, in the future, be used as a virtual sensor. Initially, a survey was made of the production history, to observe the product with the largest quality losses. From this, correlated with the critical points of measurement profile in the CD direction. Were made three artificial neural networks, Multi-Layer Perceptron (MLP) type to predict the three sectors of the profile of the paper machine. Being, the front side (FS), middle of profile (M) and drive side (DS). We found some changes in the order in which variables correlated with twist, from the sensitivity analysis, it was revealed that the most important variable and sensitive, respectively for the FS, MS and DS were total flow from the top layer headbox, vapor pressure in the 6th group and flow lateral mass of the paperboard bottom layer (Module Edge). The correlation coefficients for RNA's were 0.426, 0.557 and 0.597, respectively, in this order, for the sectors of the FS, M and DS. In conclusion, the results allow us to indicate that the models are able to represent the process. A estabilidade dimensional do papel pode sofrer alterações devido a troca de umidade do meio, liberando o stress latente adquirido no processo de fabricação. Um dos resultados dessa liberação de tensões é o encanoamento diagonal, ou também conhecido por twist. Na indústria de papel, as medições laboratoriais são importantes para garantir a qualidade final do material entregue ao cliente. Particularmente na medição do encanoamento diagonal, as amostragens são feitas no final de cada rolo jumbo, demandando tempo para eventuais correções, por parte da operação. Com essas pressuposições, este estudo tem por objetivo realizar uma análise de sensibilidade das diferentes variáveis de entrada de uma máquina industrial de papel, juntamente com algumas medições laboratoriais, relacionadas à propriedade do papel denominada de encanoamento diagonal e propor redes neurais artificiais para, no futuro, serem usadas como um sensor virtual no processo de produção industrial. Inicialmente foi feito um levantamento do histórico de produção para observar os produtos com as maiores perdas de qualidade. A partir disso, correlacionados com os pontos críticos do perfil de medição na direção CD, foram confeccionadas três redes neurais artificiais, do tipo Multi-Layer Perceptron (MLP), para predizer três setores do perfil da máquina de papel, sendo, o lado do comando (LC), meio de máquina (M) e lado do acionamento (LA). Encontramos algumas alterações na ordem em que as variáveis correlacionavam com o encanoamento diagonal. A partir da análise de sensibilidade, foi revelado, que a variável mais importante e sensível, respectivamente para o lado comando, meio e acionamento da máquina, foram fluxo total da caixa de entrada da camada cobertura, pressão de vapor no 6o grupo e fluxo de massa lateral da camada base do papel cartão (Module Edge). Os coeficientes de correlação para as RNA s foram de 0,426; 0,557 e 0,597, respectivamente, nesta ordem, para os setores do LC, M e LA. Portanto, nos permitindo concluir que os resultados indicam que os modelos são aptos para representar o processo.