Nonlinear Decision Boundaries Research Articles

CapsRule: Explainable Deep Learning for Classifying Network Attacks.

Despite the potential deep learning (DL) algorithms have shown, their lack of transparency hinders their widespread application. Extracting if-then rules from deep neural networks is a powerful explanation method to capture nonlinear local behaviors. However, existing rule extraction methods suffer from inefficiency, incomprehensibility, infidelity, and not scaling well. Concerning security applications, they are not optimized regarding the decision boundary, data types and ranges, classification tasks, and dataset size. In this article, we propose CapsRule, an effective and efficient rule-based DL explanation method dedicated to classifying network attacks. It extracts high-fidelity rules from the feed-forward capsule network that explains how an input sample is classified. Using precomputed coupling coefficients, the training phase overlaps the rule extraction process to increase efficiency. The activation vector of a capsule can represent semantic intelligence about the attributes of the input sample. The rules extracted from CapsRule address the major concerns of network attack detection. The rules: 1) approximate the nonlinear decision boundary of the underlying data; 2) reduce the number of false positives significantly; 3) increase transparency; and 4) help find errors and noise in the data. We evaluate CapsRule on the CICDDoS2019 dataset that contains over a million of the most advanced Distributed Denial-of-Service (DDoS) attacks. The extensive evaluation shows that it generates accurate, high-fidelity, and comprehensible rules. CapsRule achieves an average accuracy of 99.0% and a false positive rate of 0.70% for reflection- and exploitation-based attacks. We verify that the learned features from the rulesets match our domain-specific knowledge. They also help find flaws in the dataset generation process and erroneous patterns caused by attack simulators.

A Hybrid Algorithm for Multiple Disease Prediction: Radial Basis Function and Logistic Regression

Disease prediction is an important aspect of modern medicine, which aims to diagnose disease early and provide appropriate treatment to patients. This research uses a hybrid approach that combines the RBF (Radial Basis Function) kernel algorithm with logistic regression to predict various diseases in medical datasets. This method is intended to improve prediction performance by exploiting the advantages of each algorithm. This research uses a dataset containing medical information about several diseases collected from the Kaggle dataset. First, the RBF kernel is applied to transform the data features into a more informative, non-linear representation. Then, the logistic regression model is used to make predictions based on the features that have been processed by the RBF kernel. In this research, the hybrid RBF (Radial Basis Function) method was proven to be superior in predicting multiple diseases. This method shows the highest accuracy of 0.9460, as well as excellent precision, recall, and F1-score values of 0.8680, 0.8097, and 0.8294, respectively. The advantage of the hybrid RBF method lies in its ability to capture complex patterns in data that other methods often cannot identify, as well as its ability to handle non-linear decision boundaries, which are a common characteristic in medical datasets. Keywords: Disease prediction, Hybrid approach, RBF kernel algorithm, Logistic regression, medical datasets

Crowdsourcing with the drift diffusion model of decision making.

Crowdsourcing involves the use of annotated labels with unknown reliability to estimate ground truth labels in datasets. A common task in crowdsourcing involves estimating reliabilities of annotators (such as through the sensitivities and specificities of annotators in the binary label setting). In the literature, beta or dirichlet distributions are typically imposed as priors on annotator reliability. In this study, we investigated the use of a neuroscientifically validated model of decision making, known as the drift-diffusion model, as a prior on the annotator labeling process. Two experiments were conducted on synthetically generated data with non-linear (sinusoidal) decision boundaries. Variational inference was used to predict ground truth labels and annotator related parameters. Our method performed similarly to a state-of-the-art technique (SVGPCR) in prediction of crowdsourced data labels and prediction through a crowdsourced-generated Gaussian process classifier. By relying on a neuroscientifically validated model of decision making to model annotator behavior, our technique opens the avenue of predicting neuroscientific biomarkers of annotators, expanding the scope of what may be learnt about annotators in crowdsourcing tasks.

BMB-LIME: LIME with modeling local nonlinearity and uncertainty in explainability

The majority of eXplainable Artificial Intelligence (XAI) methods assume the local linearity of the decision boundary, leading to significant errors when dealing with non-linear local decision boundaries. Moreover, explanation methods relying on perturbation samples yield only a constant value for feature contributions. This study aims to overcome these limitations by introducing a novel explanation method that accurately captures non-linear local classification boundaries or regression splines while also quantifying the uncertainty associated with each feature contribution. Expanding upon the state-of-the-art XAI method, Local Interpretable Model-agnostic Explanations (LIME), we propose a nonlinear local explainer called BMB-LIME (Bootstrap aggregating Multivariate adaptive regression splines Bayesian LIME). BMB-LIME is constructed using weighted multivariate adaptive regression splines (MARS) with bootstrap aggregating and models the uncertainty of feature contributions within a Bayesian framework. Through a series of experiments, we demonstrate the superior performance of the BMB-LIME local explainer over baseline methods in terms of local fidelity and stability. Its adjusted r-squared is statistically significant by at least 80.5 percent, and it consistently demonstrates over 89.1 percent consistency in feature contribution and Jaccard similarity across diverse real-world and simulated datasets. The proposed method not only enhances fidelity and stability but also improves the measurement of uncertainty in explanations, thereby significantly contributing to the trustworthiness and transparency of XAI.

A deep clustering framework for load pattern segmentation

In recent years, the widespread use of smart meters in power networks has generated a wealth of data from electricity customers. However, much of this real-world smart meter dataset lacks labeled data, posing a significant challenge that can be addressed by clustering. Traditional clustering methods often struggle in high-dimensional spaces, leading to less accurate results and increased computational demands. In response to these challenges, this research introduces a framework that utilizes a deep-learning-based clustering approach to address the issue of accurate customer clustering based on usage patterns in unlabeled data. Using an autoencoder, our approach integrates dimensionality reduction and clustering into an end-to-end unsupervised learning framework. Our algorithm significantly improves load profiling by tackling challenges related to nonlinear decision boundaries at the autoencoder bottleneck. Unlike traditional approaches, we propose separating the optimization of reconstruction and cluster loss, bridging the gap between clustering quality and reconstruction efficiency. We rigorously analyze the performance of our approach by comparing classical and state-of-the-art algorithms using two real-world smart meter data. We provide a comprehensive comparative analysis of our method against five common dimension reduction techniques used in high-dimensional clustering. The experimental analysis concludes that the proposed algorithm significantly enhances load profiling more than others, as confirmed through detailed load curve analysis and clustering validity indexes. This comprehensive assessment highlights the effectiveness and versatility of our proposed methodology when compared to others. Moreover, this research advances load profiling in smart grid analytics, providing practical insights for utilities and stakeholders looking to optimize power network operations.

SEMG-Based Gesture Recognition Using Temporal History.

Surface electromyography (sEMG) patterns have been decoded using learning-based methods that determine complicated nonlinear decision boundaries. However, overlapping classes in sEMG pattern recognition still degrade the classification accuracy because they cannot be separated by the decision boundaries. We hypothesized that certain overlapping classes can be separated while tracing the temporal history of sEMG patterns. Therefore, a novel post-processing method is proposed to adjust classification errors using the separated patterns from the temporal history of overlapping classes. The proposed method confirms the confidence of the prediction result by calculating the instantaneous pattern separability for the sequential sEMG input. The prediction result with high separability pattern is considered to have a high confidence of being correct (reliable). This result is stored for adjusting the next sEMG input. When the subsequent prediction is identified as having low confidence (unreliable), the predicted result is adjusted using the stored reliable predicted results. The proposed method adds an adjustment step to an existing classifier (maximum likelihood classifier (MLC), k-nearest neighbor (KNN), and support vector machine (SVM)), such that it can be attached to the back-end regardless of the type of classifier. Ten subjects performed 13 types of hand gestures, including overlapping patterns. The overall classification accuracy was enhanced to 88.93%(+8.12%p, MLC), 91.31%(+7.68%p, KNN), and 99.65%(+11.63%p, SVM) after the implementation of the proposed post-processing. Additionally, a faster and more accurate gesture classification was achieved with accuracy enhancement before gesture completion as 85.62%(+4.23%p, MLC), 89.77%(+4.23%p, KNN), and 97.62%(+11.12%p, SVM).

Data re-uploading with a single qudit

Quantum two-level systems, i.e., qubits, form the basis for most quantum machine learning approaches that have been proposed throughout the years. However, higher dimensional quantum systems constitute a promising alternative and are increasingly explored in theory and practice. Here, we explore the capabilities of multi-level quantum systems, so-called qudits, for their use in a quantum machine learning context. We formulate classification and regression problems with the data re-uploading approach and demonstrate that a quantum circuit operating on a single qudit is able to successfully learn highly non-linear decision boundaries of classification problems such as the MNIST digit recognition problem. We demonstrate that the performance strongly depends on the relation between the qudit states representing the labels and the structure of labels in the training data set. Such a bias can lead to substantial performance improvement over qubit-based circuits in cases where the labels, the qudit states, and the operators employed to encode the data are well-aligned. Furthermore, we elucidate the influence of the choice of the elementary operators and show that a squeezing operator is necessary to achieve good performances. We also show that there exists a trade-off for qudit systems between the number of circuit-generating operators in each processing layer and the total number of layers needed to achieve a given accuracy. Finally, we compare classification results from numerically exact simulations and their equivalent implementation on actual IBM quantum hardware. The findings of our work support the notion that qudit-based algorithms exhibit attractive traits and constitute a promising route to increasing the computational capabilities of quantum machine learning approaches.

Wasserstein filter for variable screening in binary classification in the reproducing kernel Hilbert space

The aim of this paper is to develop a marginal screening method for variable screening in high-dimensional binary classification based on the Wasserstein distance accounting for the distributional difference. Many existing screening methods, such as the two-sample t-test and Kolmogorov test, have been developed under the parametric/nonparametric modeling assumptions to reduce the dimension of the predictors. However, such modeling specifications or nonparametric approaches are associated with the probability measure induced by the predictor in a Euclidean space. While many machine learning methods have successfully found the nonlinear decision boundary in the transformed space, called the reproducing kernel Hilbert space (RKHS), we consider the Wasserstein filter's capacity to detect the distributional difference between two probability measures induced by the nonlinear function of the predictor in the RKHS. Thereby, we can flexibly filter out the non-informative predictors associated with the binary classification, as well as escape the modeling assumptions required in a Euclidean space. We prove that the Wasserstein filter satisfies the sure screening property under some mild conditions. We also demonstrate the advantages of our proposed approach by comparing the finite sample performance of it with those of the existing choices through simulation studies, as well as through application to lung cancer data.

Optimizing Quantum Classification Algorithms on Classical Benchmark Datasets.

The discovery of quantum algorithms offering provable advantages over the best known classical alternatives, together with the parallel ongoing revolution brought about by classical artificial intelligence, motivates a search for applications of quantum information processing methods to machine learning. Among several proposals in this domain, quantum kernel methods have emerged as particularly promising candidates. However, while some rigorous speedups on certain highly specific problems have been formally proven, only empirical proof-of-principle results have been reported so far for real-world datasets. Moreover, no systematic procedure is known, in general, to fine tune and optimize the performances of kernel-based quantum classification algorithms. At the same time, certain limitations such as kernel concentration effects-hindering the trainability of quantum classifiers-have also been recently pointed out. In this work, we propose several general-purpose optimization methods and best practices designed to enhance the practical usefulness of fidelity-based quantum classification algorithms. Specifically, we first describe a data pre-processing strategy that, by preserving the relevant relationships between data points when processed through quantum feature maps, substantially alleviates the effect of kernel concentration on structured datasets. We also introduce a classical post-processing method that, based on standard fidelity measures estimated on a quantum processor, yields non-linear decision boundaries in the feature Hilbert space, thus achieving the quantum counterpart of the radial basis functions technique that is widely employed in classical kernel methods. Finally, we apply the so-called quantum metric learning protocol to engineer and adjust trainable quantum embeddings, demonstrating substantial performance improvements on several paradigmatic real-world classification tasks.

Comparison of linear and non-linear decision boundaries to detect feedlot bloat using intensive data collection systems on Angus × Hereford steers

Ruminal tympany (bloat) has long been an issue for large and small livestock operations. Though improvements in feedlot management practices have reduced its occurrence, it is still highly prevalent and is known to detrimentally affect animal performance, welfare, and in many instances, lead to animal death. Current decision support systems and diet formulation software omit the inclusion of bloat prediction based on animal performance. Here, we aim to predict bloat incidence in implanted and non-implanted feedlot steers from performance data comparing linear (LDB) and non-linear decision boundaries. Eighteen crossbred Angus × Hereford steers: BW (491.13 ± 25.78 kg) and age (12 ± 1 mo) were randomly distributed into implanted and non-implanted treatments. All animals were randomly assigned to one of two pens fit with automated monitoring systems for BW, freshwater intake, and water intake behavior: water intake event visit, no water intake event visit (NWIE), and time spent drinking. DM intake (DMI) was individually recorded from all animals through the Calan Gate system for 135 d (30 d adaptation, 105 d experimental diet). Incidences of bloat were recorded as bloat instances regardless of severity to ensure that early onset detection of bloat was recorded and properly identified in predictive models. Logistic regression with a binomial distribution and a logit link function was utilized to predict the incidences of bloat through LDB. Feature selection and penalization of coefficients were explored through L1 (sum of absolute values) and L2 (sum of squares) penalization to avoid overfitting of models. Additional NLDB and a non-parametric LDB are examined for prediction. Accuracy, specificity, and sensitivity were high for the models reported. No significant differences were observed between LDB and NLDB, with the highest specificity (predicting bloat) value of 0.820 for stepwise feature selection algorithms, and a value of 0.832 for the artificial neural network. Highest accuracy was 0.829 for ridge regression, and 0.847 for the random forest with hyperparameter tuning. DM intake, BW, and NWIE were the three most important variables for the prediction of feedlot bloat showing clear drops in DMI and BW and increases in NWIE when animals bloated. The lack of difference in predictive performance between LDB and NLDB highlights the often-overlooked concept that machine learning algorithms are not always the only/best modeling technique. Additionally, the models reported herein carry acceptable predictive performance for inclusion into management decisions that reduce bloat incidences in feedlot cattle.

Why did AI get this one wrong? — Tree-based explanations of machine learning model predictions

Increasingly complex learning methods such as boosting, bagging and deep learning have made ML models more accurate, but harder to interpret and explain, culminating in black-box machine learning models. Model developers and users alike are often presented with a trade-off between performance and intelligibility, especially in high-stakes applications like medicine. In the present article we propose a novel methodological approach for generating explanations for the predictions of a generic machine learning model, given a specific instance for which the prediction has been made. The method, named AraucanaXAI, is based on surrogate, locally-fitted classification and regression trees that are used to provide post-hoc explanations of the prediction of a generic machine learning model. Advantages of the proposed XAI approach include superior fidelity to the original model, ability to deal with non-linear decision boundaries, and native support to both classification and regression problems. We provide a packaged, open-source implementation of the AraucanaXAI method and evaluate its behaviour in a number of different settings that are commonly encountered in medical applications of AI. These include potential disagreement between the model prediction and physician’s expert opinion and low reliability of the prediction due to data scarcity.

Machine learning approach for hemorrhagic transformation prediction: Capturing predictors' interaction.

Patients with ischemic stroke frequently develop hemorrhagic transformation (HT), which could potentially worsen the prognosis. The objectives of the current study were to determine the incidence and predictors of HT, to evaluate predictor interaction, and to identify the optimal predicting models. A prospective study included 360 patients with ischemic stroke, of whom 354 successfully continued the study. Patients were subjected to thorough general and neurological examination and T2 diffusion-weighted MRI, at admission and 1 week later to determine the incidence of HT. HT predictors were selected by a filter-based minimum redundancy maximum relevance (mRMR) algorithm independent of model performance. Several machine learning algorithms including multivariable logistic regression classifier (LRC), support vector classifier (SVC), random forest classifier (RFC), gradient boosting classifier (GBC), and multilayer perceptron classifier (MLPC) were optimized for HT prediction in a randomly selected half of the sample (training set) and tested in the other half of the sample (testing set). The model predictive performance was evaluated using receiver operator characteristic (ROC) and visualized by observing case distribution relative to the models' predicted three-dimensional (3D) hypothesis spaces within the testing dataset true feature space. The interaction between predictors was investigated using generalized additive modeling (GAM). The incidence of HT in patients with ischemic stroke was 19.8%. Infarction size, cerebral microbleeds (CMB), and the National Institute of Health stroke scale (NIHSS) were identified as the best HT predictors. RFC (AUC: 0.91, 95% CI: 0.85-0.95) and GBC (AUC: 0.91, 95% CI: 0.86-0.95) demonstrated significantly superior performance compared to LRC (AUC: 0.85, 95% CI: 0.79-0.91) and MLPC (AUC: 0.85, 95% CI: 0.78-0.92). SVC (AUC: 0.90, 95% CI: 0.85-0.94) outperformed LRC and MLPC but did not reach statistical significance. LRC and MLPC did not show significant differences. The best models' 3D hypothesis spaces demonstrated non-linear decision boundaries suggesting an interaction between predictor variables. GAM analysis demonstrated a linear and non-linear significant interaction between NIHSS and CMB and between NIHSS and infarction size, respectively. Cerebral microbleeds, NIHSS, and infarction size were identified as HT predictors. The best predicting models were RFC and GBC capable of capturing nonlinear interaction between predictors. Predictor interaction suggests a dynamic, rather than, fixed cutoff risk value for any of these predictors.

Prediction of microbial growth via the hyperconic neural network approach

The design and control of biological reactors depends upon the correct calibration and selection of kinetic parameters from the exponential phase in the microbial growth. The kinetic microbial parameters that are reported in the literature of biochemical engineering are obtained based on specific growth conditions, which limits their potential application in a wide variety of dynamic models. In this paper, we propose an adaptive machine-learning approach to predict microbial growth, and provide information on the effect that variations of pH and concentration of the culture media have on the growth rate. We study biological reactions and obtain a set of experimental exploratory data using the Pseudomonas aeruginosa (i.e., P. aeruginosa). The versatile and robust metabolism of P. aeruginosa is responsible for its ability to grow in different environment conditions, even at low nutrient and oxygen levels, in a sample range of temperatures (4∘-42∘C) and polluted sites. Our first contribution is to propose a technique to gather experimental data via measurements of optical density, microbial growth, from the Multiskan™ FC Microplate Photometer, Thermo Scientific. Our second contribution consists of building a surrogate model by successfully adapting the Hyperconic Multilayer Perceptron (HC-MLP), a novel computational and mathematical approach, to predict microbial growth across all the set of conditions of the experimental design. HC-MLP is a state-of-the-art method that is used to define complex non-linear decision boundaries, in the parameters’ space, using a mix of ideas from conformal geometry and neural networks, and focusing on quadratic hyper-surfaces through multiple hidden layers. Consequently, we generate precise surrogates, which can even predict microbial growth in values not evaluated during the experimental stage. Finally, our statistical testing and comparisons validate that the proposed experimental, mathematical and computational framework is robust and capable of predicting the dynamic growth of bacteria P. aeruginosa using two main operation conditions: pH and concentration culture media at pH 7.0. In the future, we plan to apply our proposed methodology to other bacterial strains and advance HC-MLP for forecasting dynamics of other multiple microbial measurements under a wide variety of conditions.

Generative Adversarial Networks for Gearbox of Wind Turbine With Unbalanced Data Sets in Fault Diagnosis

Signal measurement and diagnosis of wind turbine gearbox are very important for equipment maintenance. Generative adversarial networks (GAN) are particularly outstanding in data generation due to its game mechanism. An improved gear fault diagnosis method based on GAN for unbalanced data sets is proposed in this paper. Firstly, the fault data is encoded by binary vectorization based on kurtosis perceptron. Secondly, the individuals that fit the fitness are selected by the deterministic selection and the macro factor code string is crossed by multiple points. Then, gaussian mutation is focused on searching for local fault points. Finally, the nonlinear decision boundary is established by the logistic regression auxiliary classifier. The effectiveness of the proposed method was verified by three groups of comparison experiments. Compared with the existing methods, the proposed method has a better ability for fault feature generation, classification, and diagnosis accuracy under unbalanced data sets.

Nonlinear Schrödinger Kernel for Hardware Acceleration of Machine Learning

Alternative machine learning approaches that have extremely low latency and can work with only a small training dataset are needed for applications where the insatiable demands of deep learning methods for computing power and large training data cannot be met. Here we report a new optical accelerator for AI that exploits femtosecond pulses for both data acquisition and computing enabling classification at short time scales for fast optical imaging, sensing, and metrology without increasing data dimensions. Modulation of data onto the spectrum of femtosecond optical pulses followed by projection into a new space using nonlinear optics reduces the latency in the nonlinear classification of certain data by orders of magnitude. This Femtocomputing approach is validated by the classification of various datasets, including brain intracranial pressure, cancer cell imaging, spoken digit recognition, and the classic exclusive OR benchmark for nonlinear operation. The concept is demonstrated by seeding the nonlinear effect that is responsible for many fascinating natural phenomena, such as optical rogue waves. Stimulation of nonlinear optical interactions with spectrally modulated data transforms the data such that a computationally-light linear algorithm can learn a nonlinear decision boundary that separates the data into the correct classes. Since the optical kernel is not trained, its performance is inevitably data-dependent. Quantitative comparison with a popular numerical kernel offers insights into how this physical technique accelerates inference. Single-shot operation is demonstrated using time stretch data acquisition.

Benchmark of Data Processing Methods and Machine Learning Models for Gut Microbiome-Based Diagnosis of Inflammatory Bowel Disease.

Patients with inflammatory bowel disease (IBD) wait months and undergo numerous invasive procedures between the initial appearance of symptoms and receiving a diagnosis. In order to reduce time until diagnosis and improve patient wellbeing, machine learning algorithms capable of diagnosing IBD from the gut microbiome’s composition are currently being explored. To date, these models have had limited clinical application due to decreased performance when applied to a new cohort of patient samples. Various methods have been developed to analyze microbiome data which may improve the generalizability of machine learning IBD diagnostic tests. With an abundance of methods, there is a need to benchmark the performance and generalizability of various machine learning pipelines (from data processing to training a machine learning model) for microbiome-based IBD diagnostic tools. We collected fifteen 16S rRNA microbiome datasets (7,707 samples) from North America to benchmark combinations of gut microbiome features, data normalization and transformation methods, batch effect correction methods, and machine learning models. Pipeline generalizability to new cohorts of patients was evaluated with two binary classification metrics following leave-one-dataset-out cross (LODO) validation, where all samples from one study were left out of the training set and tested upon. We demonstrate that taxonomic features processed with a compositional transformation method and batch effect correction with the naive zero-centering method attain the best classification performance. In addition, machine learning models that identify non-linear decision boundaries between labels are more generalizable than those that are linearly constrained. Lastly, we illustrate the importance of generating a curated training dataset to ensure similar performance across patient demographics. These findings will help improve the generalizability of machine learning models as we move towards non-invasive diagnostic and disease management tools for patients with IBD.

Two-stream network with phase map for few-shot classification

As an important research direction of deep learning, few-shot classification mainly solves problems of classification when training samples are scarce. Under this circumstance, how to effectively use the information of existing samples has become the key to solve few-shot classification problems. We propose a method to fully mine existing samples information, which can be implemented in the training and testing stages. Specifically, we introduce a two-stream neural network as an embedding model to extract features of the original image and its phase map, and perform fine-grained adaptive feature fusion to them. The algorithm can learn and extract more fully-informed fusion features under the network. Besides, we also introduce Circle Loss in loss function, which has a non-linear decision boundary. Under the joint supervision of Soft-max Loss and Circle Loss, the distinction between different classes of features in embedding space increases. We conduct experiments on miniImageNet, tieredImageNet, and CIFAR-FS, respectively. The accuracy rate is increased by 0.99%, 0.96%, and 0.91%, which proves the effectiveness of our algorithm.

A pilot exploration on the use of NIR monitored haemodynamics in gesture recognition for transradial prosthesis control

• Evaluation of hemodynamics as a physiological monitor for prosthesis control • Meta-heuristic feature exploration • Pilot to evaluate a potentially cheaper means of prosthesis control • Hand gesture recognition using blood flow dynamics The application of near infrared (NIR) sensing for gesture recognition carries a high appeal as it presents a haemodynamic monitoring method, which provides an alternative to the conventional electromyography sensing method for prosthesis control, which is limited by muscle fatigue and is said to be expensive depending on the specifications. Using an affordable (£25), low sampling and high resolution wearable NIR sensor, a feature selection exercise was conducted where 23 signal features were extracted from the acquired signal and downselected to an optimal 11, which allow for gesture recognition while enhancing recognition accuracy. The selected features were tested and compared with a reduced feature set with four different classifiers, namely multilayer perceptron neural network (MLPNN), Bayesian classifier (BC), linear and quadratic discriminant analysis (LDA and QDA) across four different gestures in five non-amputated participants. The best improvement in classification was produced with the MLPNN and QDA, owing to their overall model complexity and ability to separate data clusters using nonlinear decision boundaries, therein validating the candidate list of features which could be used to characterise and extract more information from haemodynamic based signals. Further work in this area can include an exercise to determine the optimal spacing of the NIR emitters and receivers to allow for maximal penetrative depth and therein an increased amount of physiological information acquired in the signal, and an investigation to observe the extent which anatomical contraction force can be differentiated using NIR.

ELECTRE tree: a machine learning approach to infer ELECTRE Tri-B parameters

PurposeThis paper presents an algorithm that can elicitate (infer) all or any combination of elimination and choice expressing reality (ELECTRE) Tri-B parameters. For example, a decision maker can maintain the values for indifference, preference and veto thresholds, and the study’s algorithm can find the criteria weights, reference profiles and the lambda cutting level. The study’s approach is inspired by a machine learning ensemble technique, the random forest, and for that, the authors named the study’s approach as ELECTRE tree algorithm.Design/methodology/approachFirst, the authors generate a set of ELECTRE Tri-B models, where each model solves a random sample of criteria and alternates. Each sample is made with replacement, having at least two criteria and between 10% and 25% of alternates. Each model has its parameters optimized by a genetic algorithm (GA) that can use an ordered cluster or an assignment example as a reference to the optimization. Finally, after the optimization phase, two procedures can be performed; the first one will merge all models, finding in this way the elicitated parameters and in the second procedure, each alternate is classified (voted) by each separated model, and the majority vote decides the final class.FindingsThe authors have noted that concerning the voting procedure, nonlinear decision boundaries are generated and they can be suitable in analyzing problems of the same nature. In contrast, the merged model generates linear decision boundaries.Originality/valueThe elicitation of ELECTRE Tri-B parameters is made by an ensemble technique that is composed of a set of multicriteria models that are engaged in generating robust solutions.

An optical neural chip for implementing complex-valued neural network

Complex-valued neural networks have many advantages over their real-valued counterparts. Conventional digital electronic computing platforms are incapable of executing truly complex-valued representations and operations. In contrast, optical computing platforms that encode information in both phase and magnitude can execute complex arithmetic by optical interference, offering significantly enhanced computational speed and energy efficiency. However, to date, most demonstrations of optical neural networks still only utilize conventional real-valued frameworks that are designed for digital computers, forfeiting many of the advantages of optical computing such as efficient complex-valued operations. In this article, we highlight an optical neural chip (ONC) that implements truly complex-valued neural networks. We benchmark the performance of our complex-valued ONC in four settings: simple Boolean tasks, species classification of an Iris dataset, classifying nonlinear datasets (Circle and Spiral), and handwriting recognition. Strong learning capabilities (i.e., high accuracy, fast convergence and the capability to construct nonlinear decision boundaries) are achieved by our complex-valued ONC compared to its real-valued counterpart.