Observability constrains expertise-dependent kinematic readout in action prediction.

Condition monitoring of track geometry irregularities from onboard measurements is a cost-effective method for daily surveillance of track quality. The monitoring of Alignment Level (AL) and Cross Level (CL) track irregularities is challenging due to the nonlinearities of the contact between wheels and rails. Recently, the authors proposed a signal-based method in combination with a machine learning (ML) fault classifier to monitor AL and CL track irregularities based on bogie frame accelerations. The authors concluded that the Support Vector Machine (SVM) fault classifier outperformed other traditional ML classifiers. Thus, an important question arises: Is the previously reported decision boundary an optimal boundary? The objective of this research investigation is to obtain an optimal decision boundary according to theory of probabilistic classification and compare the same against the SVM decision boundary. In this investigation, the classifiers are trained with results of numerical simulations and validated with measurements acquired by a diagnostic vehicle on straight track sections of a high-speed line (300 km/h). A fault classifier based on Maximum A Posterior Naïve Bayes (MAP-NB) classification is developed. It is shown that the MAP-NB classifier generates an optimal decision boundary and outperforms other classifiers in the validation phase with classification accuracy of 95.9 ± 0.2%and kappa value of 80.4 ± 0.6%. Moreover, the Linear SVM (L SVM) and Gaussian-SVM (G SVM) classifiers give similar performance with slightly lower accuracy and kappa value. The decision boundaries of previously reported SVM based fault classifiers are very close to the optimal MAP-NB decision boundary. Thus, this further strengthens the idea of implementing statistical fault classifiers to monitor the track irregularities based on dynamics in the lateral plane via in-service vehicles. The proposed method contributes towards digitalization of rail networks through condition-based and predictive maintenance.KeywordsHigh-speed railwayLateral dynamicsOnboard diagnosticsProbabilistic fault classifier

The most widely used account of decision-making proposes that people choose between alternatives by accumulating evidence in favor of each alternative until this evidence reaches a decision boundary. It is frequently assumed that this decision boundary stays constant during a decision, depending on the evidence collected but not on time. Recent experimental and theoretical work has challenged this assumption, showing that constant decision boundaries are, in some circumstances, sub-optimal. We introduce a theoretical model that facilitates identification of the optimal decision boundaries under a wide range of conditions. Time-varying optimal decision boundaries for our model are a result only of uncertainty over the difficulty of each trial and do not require decision deadlines or costs associated with collecting evidence, as assumed by previous authors. Furthermore, the shape of optimal decision boundaries depends on the difficulties of different decisions. When some trials are very difficult, optimal boundaries decrease with time, but for tasks that only include a mixture of easy and medium difficulty trials, the optimal boundaries increase or stay constant. We also show how this simple model can be extended to more complex decision-making tasks such as when people have unequal priors or when they can choose to opt out of decisions. The theoretical model presented here provides an important framework to understand how, why, and whether decision boundaries should change over time in experiments on decision-making.

Decision letter: Neural markers of predictive coding under perceptual uncertainty revealed with Hierarchical Frequency Tagging

This work proposes a new face detection system using quantum-inspired evolutionary algorithm (QEA). The proposed detection system is based on elliptical blobs and principal component analysis (PCA). The elliptical blobs in the directional image are used to find the face candidate regions, and then PCA and QEA are employed to verify faces. Although PCA related algorithms have shown outstanding performance, there still exist some problems such as optimal decision boundary or learning capabilities. By PCA, we can obtain the optimal basis but they may not be the optimal ones for discriminating faces from non-faces. Moreover, a threshold value should be selected properly considering the success rate and false alarm rate. To solve these problems, QEA is employed to find out the optimal decision boundary under the predetermined threshold value which distinguishes between face images and non-face images. The proposed system provides learning capability by reconstructing the training database, which means that system performance can be improved as failure trials occur.

A novel nonlinear discriminant analysis method, Kernelized Decision Boundary Analysis (KDBA), is proposed in our paper, whose Decision Boundary feature vectors are the normal vector of the optimal Decision Boundary in terms of the Structure Risk Minimization principle. We also use a simple method to prove a property of Support Vector Machine (SVM) algorithm, which is combined with the optimal Decision Boundary Feature matrix to make our method consistent with the Kernel Fisher method(KFD). Moreover, KDBA is easily used in its applications, and the traditional Decision Boundary Analysis implementations are computationally expensive and sensitive to the size of the problem. Text classification problem is first used to testify the effectiveness of KDBA. Then experiments on the large-scale face database, the CAS-PEAL database, have illustrated its excellent performance compared with some popular face recognition methods such as Eigenface, Fisherface, and KFD.

Predicting intentions from observing another agent’s behaviours is often thought to depend on motor resonance – i.e., the motor system’s response to a perceived movement by the activation of its stored motor counterpart, but observers might also rely on prior expectations, especially when actions take place in perceptually uncertain situations. Here we assessed motor resonance during an action prediction task using transcranial magnetic stimulation to probe corticospinal excitability (CSE) and report that experimentally-induced updates in observers’ prior expectations modulate CSE when predictions are made under situations of perceptual uncertainty. We show that prior expectations are updated on the basis of both biomechanical and probabilistic prior information and that the magnitude of the CSE modulation observed across participants is explained by the magnitude of change in their prior expectations. These findings provide the first evidence that when observers predict others’ intentions, motor resonance mechanisms adapt to changes in their prior expectations. We propose that this adaptive adjustment might reflect a regulatory control mechanism that shares some similarities with that observed during action selection. Such a mechanism could help arbitrate the competition between biomechanical and probabilistic prior information when appropriate for prediction.

In this paper, the optimum decision boundaries for (N, M) differential amplitude phase-shift keying on the Rayleigh-fading channel are analyzed. A postdetection maximal ratio combining (MRC) and weighted maximal ratio combining (WMRC) diversity receivers are proposed. In the Rayleigh-fading channel, assuming a high signal-to-noise ratio and a small normalized Doppler frequency, the analytical optimum decision boundaries are obtained. In addition, it is shown that an outer optimum decision boundary is the inverse of the inner optimum decision boundary. In the proposed MRC receiver, the decision at each branch is made based on the minimum distance criterion. The performance of the MRC receiver is analyzed, in terms of the union bound for bit error probability. The proposed WMRC receiver assigns weighting factors to the decision variable at each branch, based on the optimum decision boundaries. The performance of the WMRC is investigated through computer simulation and compared with those of MRC and equal gain combining (EGC). From the results, the performances of MRC and WMRC are found to be better than those of the EGC receiver on both the Rayleigh- and Rician-fading channels. It is also found that the performance improvement of WMRC over MRC is more pronounced as the number of diversity branches increases.

An adaptive online learning method is presented to facilitate pattern classification using active sampling to identify optimal decision boundary for a stochastic oracle with a minimum number of training samples. The strategy of sampling at the current estimate of the decision boundary is shown to be optimal in the sense that the probability of convergence toward the true decision boundary at each step is maximized, offering theoretical justification on the popular strategy of category boundary sampling used by many query learning algorithms.

In this paper, we propose a new decision method based on a support vector machine for quadrature amplitude modulation (QAM) to improve the bit error rate (BER) performance of QAM for arbitrary complex impairments when both modeled and unknown impairments exist simultaneously. To do this, we first propose an algorithm for assigning bit decision boundaries corresponding to each quadrant for <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">M</i> -ary QAM, considering the asymmetry of the QAM constellation caused by the impairments. Then, we determine the optimal decision boundaries by learning a linear classification model for each bit decision region from I and Q values of received QAM signals as two-dimensional support vectors. Finally, the decision rules for each bit are presented based on the obtained optimal decision boundaries, and the bit decisions are performed using the decision rules. Based on the above process, we can effectively demodulate QAM signals for arbitrary complex impairments without compensation processes. Through computer simulations, we show the proposed method outperforms conventional model-based compensation schemes in BER performance under the condition of complex impairments.

-one-class support vector classification (-OCSVC) has garnered significant attention for its remarkable performance in handling single-class classification and anomaly detection. Nonetheless, the model does not yield a unique decision boundary, and potentially compromises learning performance when the training data is contaminated by some outliers or mislabeled observations. This paper presents a novel C-parameter version of bounded one-class support vector classification (C-BOCSVC) to determine a unique decision boundary. The distance from the origin to decision boundary is the geometrical margin in the space (higher 1-dimension than feature space), and its maximization corresponds to the structural risk minimization (SRM) principle inscribed by an -norm regularization term both on the normal direction and bias of the decision boundary. To enhance the anti-noise and anti-outlier abilities of C-BOCSVC, the alternative robust version (C-RBOCSVC) is also developed, which incorporates the k-nearest neighbor relative density to assign varying weights to observations and mitigate the negative impact of outliers on the optimal decision boundary. The theoretical properties of the proposed method are successively derived, including the relationship between the solutions to the primal and dual problems, the connections between our C-BOCSVC and -OCSVC and the computational complexity. Experimental results over massive datasets demonstrate the feasibility and reliability of our C-BOCSVC, and highlight the superior performance of C-RBOCSVC compared to other state-of-the-art one-class classifiers when data is contaminated. The demo code of this work is publicly available at https://github.com/Zhangmath1122/C-RBOCSVC.

Manual labelling of changes present in a pair of remotely sensed hyperspectral images is costly and time-consuming. As the label information is less, one might take an active learning approach where the machine learning model can learn with smart human supervision. However, there is a lack of research in the literature around change detection in partially labelled hyperspectral images. This article proposes a convolutional autoencoder-based model for detecting changes in hyperspectral images, which would reduce the data’s dimensionality and learn from unlabelled samples. The final classifier model has been re-trained using active learning. After each epoch, the model builds a decision boundary and automatically picks samples for manual labelling based on an uncertainty parameter modelled using the beta distribution function. The selected pixels’ label information is fed into the model to improve the accuracy of change detection, and the model is iterated a number of times by adding the labelled examples to the training set. Starting with a small and non-optimal training set, the model is permitted to query for the labels of k most uncertain samples at each iteration to build the updated decision boundary. It has been seen that the optimal decision boundary could be constructed by fewer labelled samples only and thus eliminating the requirement for a huge training set. According to the results, the suggested model needs extremely minimal training data (only 17.14 % and 18.57 % of training data for Bay Area and Santa Barbara images respectively) to obtain a comparatively higher level of performance.

We introduce two semi-supervised models for the classification of remote sensing image data. The models are built upon the framework of Virtual Support Vector Machines (VSVM). Generally, VSVM follow a two-step learning procedure: A Support Vector Machines (SVM) model is learned to determine and extract labeled samples that constitute the decision boundary with the maximum margin between thematic classes, i.e., the Support Vectors (SVs). The SVs govern the creation of so-called virtual samples. This is done by modifying, i.e., perturbing, the image features to which a decision boundary needs to be invariant. Subsequently, the classification model is learned for a second time by using the newly created virtual samples in addition to the SVs to eventually find a new optimal decision boundary. Here, we extend this concept by (i) integrating a constrained set of semi-labeled samples when establishing the final model. Thereby, the model constrainment, i.e., the selection mechanism for including solely informative semi-labeled samples, is built upon a self-learning procedure composed of two active learning heuristics. Additionally, (ii) we consecutively deploy semi-labeled samples for the creation of semi-labeled virtual samples by modifying the image features of semi-labeled samples that have become semi-labeled SVs after an initial model run. We present experimental results from classifying two multispectral data sets with a sub-meter geometric resolution. The proposed semi-supervised VSVM models exhibit the most favorable performance compared to related SVM and VSVM-based approaches, as well as (semi-)supervised CNNs, in situations with a very limited amount of available prior knowledge, i.e., labeled samples.

The application of neural classifiers for providing optimal decision boundaries of a warped and clustered M-QAM constellation affected by nonlinearity is analyzed in this paper. The classifier behavior, for the specific application, has been evaluated both by the carrier to noise ratio (CNR) degradation (DeltaC/N) due to nonlinearity for a target error rate P(c)=10(-3), and more thoroughly by classical figures of merit of the pattern recognition theory such as classification confidence and generalization capability. The influence of the probability distribution of the training examples and the effects of activation functions' sharpness (namely the temperature of the net) have also been investigated. The results, obtained on a simulation basis, indicate optimal matching with respect to upper bounds evaluated with some minor simplifying hypothesis, even if the overall method's effectiveness can be adequate only for mild nonlinearity conditions.

A variety of modifications have been employed to learning vector quantization (LVQ) algorithms using either crisp or soft windows for selection of data. Although these schemes have been shown in practice to improve performance, a theoretical study on the influence of windows has so far been limited. Here we rigorously analyze the influence of windows in a controlled environment of gaussian mixtures in high dimensions. Concepts from statistical physics and the theory of online learning allow an exact description of the training dynamics, yielding typical learning curves, convergence properties, and achievable generalization abilities. We compare the performance and demonstrate the advantages of various algorithms, including LVQ 2.1, generalized LVQ (GLVQ), Learning from Mistakes (LFM) and Robust Soft LVQ (RSLVQ). We find that the selection of the window parameter highly influences the learning curves but not, surprisingly, the asymptotic performances of LVQ 2.1 and RSLVQ. Although the prototypes of LVQ 2.1 exhibit divergent behavior, the resulting decision boundary coincides with the optimal decision boundary, thus yielding optimal generalization ability.

Deep learning models have achieved state of the art performances, especially for computer vision applications. Much of the recent successes can be attributed to the existence of large, high quality, labeled datasets. However, in many real-world applications, collecting similar datasets is often cumbersome and time consuming. For instance, developing robust automatic target recognition models from infrared images still faces major challenges. This is mainly due to the difficulty of acquiring high resolution inputs, sensitivity to the thermal sensors' calibration, meteorological conditions, targets' scale and viewpoint invariance. Ideally, a good training set should contain enough variations within each class for the model to learn the most optimal decision boundaries. However, when there are under-represented regions in the training feature space, especially in low data regime or in presence of low-quality inputs, the model risks learning sub-optimal decision boundaries, resulting in sub-optimal predictions. This dissertation presents novel data augmentation (DA) strategies aimed at improving the performance of machine learning models in low data regimes. The proposed techniques are designed to augment limited labeled datasets, providing the models with additional information to learn from.\\ The first contribution of this work is the development of Confidence-Guided Generative Augmentation (CGG-DA), a technique that trains and learns a generative model, such as Variational Autoencoder (VAE) and Deep Convolutional Generative Adversarial Networks (DCGAN), to generate synthetic augmentations. These generative models can generate labeled and/or unlabeled data by drawing from the same distribution as the under-performing samples based on a baseline reference model. By augmenting the training dataset with these synthetic images, CGG-DA aims to bridge the performance gap across different regions of the training feature space. We also introduce a Tool-Supported Contextual Augmentation (TSC-DA) technique that leverages existing ML models, such as classifiers or object detectors, to label available unlabeled data. Samples with consistent and high confidence predictions are used as labeled augmentations. On the other hand, samples with low confidence predictions might still contain some information even though they are more likely to be noisy and inconsistent. Hence, we keep them and use them as unlabeled samples during. Our third proposed DA explores the use of existing ML tools and external image repositories for data augmentation. This approach, called Guided External Data Augmentation (EG-DA), leverages external image repositories to augment the available dataset. External repositories are typically noisy, and might include a lot of out-of-distribution (OOD) samples. If included in the training process without proper handling,