Low self-esteem and interpersonal needs (i.e., thwarted belongingness (TB) and perceived burdensomeness (PB)) have a major impact on depression and suicide attempts. Individuals seek social connectedness on social media to boost and alleviate their loneliness. Social media platforms allow people to express their thoughts, experiences, beliefs, and emotions. Prior studies on mental health from social media have focused on symptoms, causes, and disorders. Whereas an initial screening of social media content for interpersonal risk factors and low self-esteem may raise early alerts and assign therapists to at-risk users of mental disturbance. Standardized scales measure self-esteem and interpersonal needs from questions created using psychological theories. In the current research, we introduce a psychology-grounded and expertly annotated dataset, LoST: Low Self esTeem, to study and detect low self-esteem on Reddit. Through an annotation approach involving checks on coherence, correctness, consistency, and reliability, we ensure gold-standard for supervised learning. We present results from different deep language models tested using two data augmentation techniques. Our findings suggest developing a class of language models that infuses psychological and clinical knowledge.
Existing neurostimulation systems implanted for the treatment of neurodegenerative disorders generally deliver invariable therapy parameters, regardless of phase of the sleep/wake cycle. However, there is considerable evidence that brain activity in these conditions varies according to this cycle, with discrete patterns of dysfunction linked to loss of circadian rhythmicity, worse clinical outcomes and impaired patient quality of life. We present a targeted concept of circadian neuromodulation using a novel device platform. This system utilises stimulation of circuits important in sleep and wake regulation, delivering bioelectronic cues (Zeitgebers) aimed at entraining rhythms to more physiological patterns in a personalised and fully configurable manner. Preliminary evidence from its first use in a clinical trial setting, with brainstem arousal circuits as a surgical target, further supports its promising impact on sleep/wake pathology. Data included in this paper highlight its versatility and effectiveness on two different patient phenotypes. In addition to exploring acute and long-term electrophysiological and behavioural effects, we also discuss current caveats and future feature improvements of our proposed system, as well as its potential applicability in modifying disease progression in future therapies.
Autonomous vehicles, commonly known as self-driving cars, are rapidly gaining popularity due to their numerous advantages, such as reducing traffic, pollution, and emissions while increasing safety, convenience, and transportation connectivity. In order to accurately track the motion signal, these vehicles are now utilising advanced control techniques, such as model predictive control (MPC). However, the efficiency of MPCs heavily relies on properly tuning their weights. The primary function of the MPC is to recalculate the optimal values for the vehicle control commands, such as desired speed, steering angle, etc., while considering the dynamic model of the autonomous vehicle. The existing linear MPC models cannot reach higher efficiency because of using fixed weights without considering the error. This paper introduces a novel approach for developing an MPC model with a time-varying weights algorithm for autonomous vehicles. The study aims to minimise motion tracking errors such as lateral position and yaw angle errors. Relevant MPC weights are calculated online using fuzzy logic-based units considering the lateral position and yaw angle errors. The proposed linear time-varying MPC was designed and developed using MATLAB software, resulting in improved motion tracking performance with 31.62% and 20.89% reduction of the root means square error of lateral position and yaw angle.
Recent advances in language modelling has significantly decreased the need of labelled data in text classification tasks. Transformer-based models, pre-trained on unlabeled data, can outmatch the performance of models trained from scratch for each task. However, the amount of labelled data need to fine-tune such type of model is still considerably high for domains requiring expert-level annotators, like the legal domain. This paper investigates the best strategies for optimizing the use of a small labeled dataset and large amounts of unlabeled data and perform a classification task in the legal area with 50 predefined topics. More specifically, we use the records of demands to a Brazilian Public Prosecutor's Office aiming to assign the descriptions in one of the subjects, which currently demands deep legal knowledge for manual filling. The task of optimizing the performance of classifiers in this scenario is especially challenging, given the low amount of resources available regarding the Portuguese language, especially in the legal domain. Our results demonstrate that classic supervised models such as logistic regression and SVM and the ensembles random forest and gradient boosting achieve better performance along with embeddings extracted with word2vec when compared to BERT language model. The latter demonstrates superior performance in association with the architecture of the model itself as a classifier, having surpassed all previous models in that regard. The best result was obtained with Unsupervised Data Augmentation (UDA), which jointly uses BERT, data augmentation, and strategies of semi-supervised learning, with an accuracy of 80.7% in the aforementioned task.
Autonomous transportation systems such as road vehicles or vessels require the consideration of the static and dynamic environment to dislocate without collision. Anticipating the behavior of an agent in a given situation is required to adequately react to it in time. Developing deep learning-based models has become the dominant approach to motion prediction recently. The social environment is often considered through a CNN-LSTM-based sub-module processing a $\textit{social tensor}$ that includes information of the past trajectory of surrounding agents. For the proposed transformer-based trajectory prediction model, an alternative, computationally more efficient social tensor definition and processing is suggested. It considers the interdependencies between target and surrounding agents at each time step directly instead of relying on information of last hidden LSTM states of individually processed agents. A transformer-based sub-module, the Social Tensor Transformer, is integrated into the overall prediction model. It is responsible for enriching the target agent's dislocation features with social interaction information obtained from the social tensor. For the awareness of spatial limitations, dislocation features are defined in relation to the navigable area. This replaces additional, computationally expensive map processing sub-modules. An ablation study shows, that for longer prediction horizons, the deviation of the predicted trajectory from the ground truth is lower compared to a spatially and socially agnostic model. Even if the performance gain from a spatial-only to a spatial and social context-sensitive model is small in terms of common error measures, by visualizing the results it can be shown that the proposed model in fact is able to predict reactions to surrounding agents and explicitely allows an interpretable behavior.
Engineering support for sports has the potential to improve training performance. Most research on sports training support has focused on support in areas that directly affect the point of force action. On the other hand, even if sports with hand-held tools, the importance of lower body motor function has been indicated. This suggests that supporting areas away from the point of force action can improve the performance of novice players. Therefore, in this study, we developed an assist suit that facilitates lower body twisting for table tennis beginners during forehand swing. U sing this assistive suit, experiments were conducted for the experimental group under the following four conditions (1) “No wear (before),” (2) “Without assist,” (3) “With assist,” and (4) “No wear (after)” to clarify the effect of wearing, assist, and training. In addition, a control group in which participants hit the ball without the assist suit was set up to examine the training effect of repetition. The results showed that both the amount of waist rotation and racket velocity increased significantly in the experimental group compared to the control group, which indicates the effect of the assist suit on training.
One of the purposes of smart agriculture is to predict the yield of paddy rice using agricultural data using machine learning. LightGBM, one of the machine learning algorithms, is applied to the yield prediction problem of paddy rice in this paper. Since LightGBM has a large number of hyperparameters, the hyperparameter optimization using the stochastic schemata exploiter (SSE) is used. From the results of comparison with Genetic Algorithm (GA), it is confirmed that SSE has a fast convergence speed. In addition, it is found that the higher the mutation rate of SSE, the more converged to the global optimal solution without falling into the local solution.
Flapping-wing flying robots, as a newly emerging research hotspot, have attracted more and more researchers' attention. Compared with traditional aircraft, flapping-wing flying robots have the characteristics of high flight efficiency, good concealment, and have a wide range of application prospects. As an important power mechanism of aircraft, the research of wing is very important. In this paper, we design a wing structure that can realize the active bending of wings, which can well imitate the bending pattern of wings of birds in the natural flight process. At the same time, a wind tunnel test was carried out to measure the lift resistance of the single wing and the folded wing under the same power. The results show that the folded wing has higher flight efficiency under the same power.
Invasive Brain Computer Interface (BCI) systems through Electrocorticographic (ECoG) signals require efficient recognition of spatiotemporal patterns from a multi-electrodes sensor array. Such signals are excellent candidates for automated pattern recognition through machine learning algorithms. The importance of these patterns can be highlighted through feature extraction techniques. However, the signal variability due to non-stationarity is ignored while extracting features, and which features to use can be challenging to figure out by visual inspection. In this study, we introduce the signal split parameter to account for the variability of the signal and increase the accuracy of the machine learning classifier. We use genetic selection, which allows the selection of the optimal combination of features from a pool of 8 different feature sets. Genetic selection of features increases accuracy and reduces the BCI prediction time. Along with Genetic selection, we also use a reduced signal length, which leads to a higher Information Transfer Rate. Thus this approach enables the design of a fast and accurate motor-related EcoG BCI system.
Cities have undergone significant changes due to the rapid increase in urban population, heightened demand for resources, and growing concerns over climate change. To address these challenges, digital transformation has become a necessity. Recent advancements in Artificial Intelligence (AI) and sensing techniques, such as synthetic sensing, can elevate Digital Twins (DTs) from digital copies of physical objects to effective and efficient platforms for data collection and in-situ processing. In such a scenario, this paper presents a compre-hensive approach for developing a Traffic Monitoring System (TMS) based on Edge Intelligence (EI), specifically designed for smart cities. Our approach prioritizes the placement of intelligence as close as possible to data sources, and leverages an “opportunistic” interpretation of DT (ODT), resulting in a novel and interdisciplinary strategy to re-engineering large-scale distributed smart systems. The preliminary results of the proposed system have shown that moving computation to the edge of the network provides several benefits, including (i) enhanced inference performance, (ii) reduced bandwidth and power consumption, (iii) and decreased latencies with respect to the classic cloud -centric approach.
A supervised feature selection method selects an appropriate but concise set of features to differentiate classes, which is highly expensive for large-scale datasets. Therefore, feature selection should aim at both minimizing the number of selected features and maximizing the accuracy of classification, or any other task. However, this crucial task is computationally highly demanding on many real-world datasets and requires a very efficient algorithm to reach a set of optimal features with a limited number of fitness evaluations. For this purpose, we have proposed the binary multi-objective coordinate search (MOCS) algorithm to solve large-scale feature selection problems. To the best of our knowledge, the proposed algorithm in this paper is the first multi-objective coordinate search algorithm. In this method, we generate new individuals by flipping a variable of the candidate solutions on the Pareto front. This enables us to investigate the effectiveness of each feature in the corresponding subset. In fact, this strategy can play the role of crossover and mutation operators to generate distinct subsets of features. The reported results indicate the significant superiority of our method over NSGA-II, on five real-world large-scale datasets, particularly when the computing budget is limited. Moreover, this simple hyper-parameter-free algorithm can solve feature selection much faster and more efficiently than NSGA-II.
Feature selection is an expensive challenging task in machine learning and data mining aimed at removing irrelevant and redundant features. This contributes to an improvement in classification accuracy, as well as the budget and memory requirements for classification, or any other post-processing task conducted after feature selection. In this regard, we define feature selection as a multi-objective binary optimization task with the objectives of maximizing classification accuracy and minimizing the number of selected features. In order to select optimal features, we have proposed a binary Compact NSGA-II (CNSGA-II) algorithm. Compactness represents the population as a probability distribution to enhance evolutionary algorithms not only to be more memory-efficient but also to reduce the number of fitness evaluations. Instead of holding two populations during the optimization process, our proposed method uses several Probability Vectors (PVs) to generate new individuals. Each PV efficiently explores a region of the search space to find non-dominated solutions instead of generating candidate solutions from a small population as is the common approach in most evolutionary algorithms. To the best of our knowledge, this is the first compact multi-objective algorithm proposed for feature selection. The reported results for expensive optimization cases with a limited budget on five datasets show that the CNSGA-II performs more efficiently than the well-known NSGA-II method in terms of the hypervolume (HV) performance metric requiring less memory. The proposed method and experimental results are explained and analyzed in detail.
While advanced classifiers have been increasingly used in real-world safety-critical applications, how to properly evaluate the black-box models given specific human values remains a concern in the community. Such human values include punishing error cases of different severity in varying degrees and making compromises in general performance to reduce specific dangerous cases. In this paper, we propose a novel evaluation measure named Meta Pattern Concern Score based on the abstract representation of probabilistic prediction and the adjustable threshold for the concession in prediction confidence, to introduce the human values into multi-classifiers. Technically, we learn from the advantages and disadvantages of two kinds of common metrics, namely the confusion matrix-based evaluation measures and the loss values, so that our measure is effective as them even under general tasks, and the cross entropy loss becomes a special case of our measure in the limit. Besides, our measure can also be used to refine the model training by dynamically adjusting the learning rate. The experiments on four kinds of models and six datasets confirm the effectiveness and efficiency of our measure. And a case study shows it can not only find the ideal model reducing 0.53% of dangerous cases by only sacrificing 0.04% of training accuracy, but also refine the learning rate to train a new model averagely outperforming the original one with a 1.62% lower value of itself and 0.36% fewer number of dangerous cases.
How would considering the aesthetics of text layout, particularly line spacing with the Golden Ratio (GR) parameter, influence the ease of reading and retention? To answer this question we introduced a novel method by employing a wearable eye tracker instead of a table-mounted eye tracker. Because of their lower price, portability, and growing market, as well as convenience to setup, such head-mounted devices are suitable use for ubiquitous reading behavior analysis. For mapping gaze information from captured video scenes to the original document, we adopted the Locally Likely Arrangement Hashing (LLAH) method for robust document retrieval. In our experimental system, participants read digital documents from the screen with and without the aesthetic GR parameter. Then, gaze data captured by the eye tracker's embedded camera is mapped to the corresponding original document. Finally, our gaze analysis system extracts the intended information for statistical evaluation. Regarding to the results, significant differences in reading performance were found in the documents with and without the GR aesthetic parameter for line spacing.
Sleep Stage Classification (SSC) is a labor-intensive task, requiring experts to examine hours of electrophysiological recordings for manual classification. This is a limiting factor when it comes to leveraging sleep stages for therapeutic purposes. With increasing affordability and expansion of wearable devices, automating SSC may enable deployment of sleep-based therapies at scale. Deep Learning has gained increasing attention as a potential method to automate this process. Previous research has shown accuracy comparable to manual expert scores. However, previous approaches require sizable amount of memory and computational resources. This constrains the ability to classify in real time and deploy models on the edge. To address this gap, we aim to provide a model capable of predicting sleep stages in real-time, without requiring access to external computational sources (e.g., mobile phone, cloud). The algorithm is power efficient to enable use on embedded battery powered systems. Our compact sleep stage classifier can be deployed on most off-the-shelf microcontrollers (MCU) with constrained hardware settings. This is due to the memory footprint of our approach requiring significantly fewer operations. The model was tested on three publicly available data bases and achieved performance comparable to the state of the art, whilst reducing model complexity by orders of magnitude (up to 280 times smaller compared to state of the art). We further optimized the model with quantization of parameters to 8 bits with only an average drop of 0.95% in accuracy. When implemented in firmware, the quantized model achieves a latency of 1.6 seconds on an Arm Cortex-M4 processor, allowing its use for on-line SSC-based therapies.
This study proposes a simplified telepresence system based on extended reality for increasing the number of remote-able tasks. The system provides an immersive virtual environment that duplicates remote real spaces to enable communication among system users in different geographic locations through their digital avatars. The motion of the avatars corresponds with that of the mobile devices manipulated by the system users, and the avatars' position and orientation in the real space represent the viewpoint and viewing angle of the users. The three experiments showed that the proposed system creates a high-fidelity immersive environment in a short period of time and positively supports users' remote collaboration in the virtualized real space. The system successfully demonstrated its potential to contribute to the expansion of remotely enabled operations.
In this paper, we present a comprehensive approach for designing and analyzing control systems with minmax constraint controllers and machine learning-based virtual sensors. By leveraging the Standard Nonlinear Operator Form (SNOF), we establish the necessary conditions for global asymptotic stability and demonstrate their applicability through an illustrative example based on a modified plant model from the literature. The proposed methodology effectively handles non-linearities and constraints, ensuring stability while providing a systematic procedure for constructing a well-formed SNOF by integrating the plant, virtual sensor, and controller. The successful application of this method in the example highlights its potential for addressing complex control problems involving min-max constraints and virtual sensors in real-world scenarios. This paper contributes to the growing body of knowledge in this area and sets the stage for future advancements, including the exploration of transforming other machine learning architectures into the SNOF and extending the stability analysis to accommodate different types of nonlinearities and constraints.
In today's era of automation, mobile robots are being used for collecting meaningful information about an ambient phenomenon such as temperature or moisture distribution in an agricultural field. Most of the studies in the literature assume that the underlying information field is Gaussian, and therefore, Gaussian Process (GP)-based models are extremely popular. Furthermore, we have found that due to the inherent computational complexity of such naive GP-based techniques, most studies in the literature do not scale well beyond small-size environments, i.e., where the number of informative points $n < 1000$ . These render such a predictive model more or less useless in many practical applications. In this paper, we posit that a different technique, Generative Adversarial Network-based inpainting, for robotic information gathering can be useful. The state-of-art inpainting techniques 1) do not assume that the underlying data is Gaussian, and 2) easily scale to $n\gg 1000$ . Thus, they eliminate the two bottlenecks posed by the GP-based solutions. We have tested our hypothesis on a synthetic and a real-world crop dataset. Results show that while the inpainting technique easily scales to $1024\times 1024$ , GP-based predictions cannot. On the other hand, their solution qualities are shown to be comparable.
The detection of malicious Android applications is a major security challenge. A number of machine learning-based techniques have been put forth, and some of them have attained great accuracy. However, the diversity of apps and frequency at which new malware families are found means that the issue remains unresolved. In this paper, we use both static, dynamic and hybrid analysis to automatically classify Android apps as benign or infected. We compare all three approaches on a common dataset — the TwinDroid dataset which contains over 15,000 system call traces from over 9,000 benign and infected app. This method allows comparison on equal footing. We make further contributions on the topic of feature selection and trace abstraction.
This paper proposes a time-efficient model for solving the Weapon-target assignment (WTA) problem with actor-critic reinforcement learning. While typical heuristic algorithms and recently studied artificial neural network methodologies have shown good performance results, the previous approach has not been time-efficient in large-scale WTA problems. This paper utilizes the actor-critic framework to resolve the WTA problem, and this framework enables retrieving solutions 23 times faster than the previous deep Q-network approach. Additionally, we incorporate a recurrent neural network model of gated recurrent units (GRU) to allow agents to learn the latent state-space of the WTA problem. Our experiments demonstrate the solution quality and the time efficiency compared to traditional heuristic methods as well as recent DQN-based RL models.
