RESUMO
The study of Domain Generalization (DG) has gained considerable momentum in the Machine Learning (ML) field. Human Activity Recognition (HAR) inherently encompasses diverse domains (e.g., users, devices, or datasets), rendering it an ideal testbed for exploring Domain Generalization. Building upon recent work, this paper investigates the application of regularization methods to bridge the generalization gap between traditional models based on handcrafted features and deep neural networks. We apply various regularizers, including sparse training, Mixup, Distributionally Robust Optimization (DRO), and Sharpness-Aware Minimization (SAM), to deep learning models and assess their performance in Out-of-Distribution (OOD) settings across multiple domains using homogenized public datasets. Our results show that Mixup and SAM are the best-performing regularizers. However, they are unable to match the performance of models based on handcrafted features. This suggests that while regularization techniques can improve OOD robustness to some extent, handcrafted features remain superior for domain generalization in HAR tasks.
Assuntos
Aprendizado de Máquina , Redes Neurais de Computação , Humanos , Atividades Humanas , Conscientização , AcelerometriaRESUMO
Human Activity Recognition (HAR) has been studied extensively, yet current approaches are not capable of generalizing across different domains (i.e., subjects, devices, or datasets) with acceptable performance. This lack of generalization hinders the applicability of these models in real-world environments. As deep neural networks are becoming increasingly popular in recent work, there is a need for an explicit comparison between handcrafted and deep representations in Out-of-Distribution (OOD) settings. This paper compares both approaches in multiple domains using homogenized public datasets. First, we compare several metrics to validate three different OOD settings. In our main experiments, we then verify that even though deep learning initially outperforms models with handcrafted features, the situation is reversed as the distance from the training distribution increases. These findings support the hypothesis that handcrafted features may generalize better across specific domains.
Assuntos
Atividades Humanas , Redes Neurais de Computação , Humanos , Reconhecimento PsicológicoRESUMO
With the fast increase in the demand for location-based services and the proliferation of smartphones, the topic of indoor localization is attracting great interest. In indoor environments, users' performed activities carry useful semantic information. These activities can then be used by indoor localization systems to confirm users' current relative locations in a building. In this paper, we propose a deep-learning model based on a Convolutional Long Short-Term Memory (ConvLSTM) network to classify human activities within the indoor localization scenario using smartphone inertial sensor data. Results show that the proposed human activity recognition (HAR) model accurately identifies nine types of activities: not moving, walking, running, going up in an elevator, going down in an elevator, walking upstairs, walking downstairs, or going up and down a ramp. Moreover, predicted human activities were integrated within an existing indoor positioning system and evaluated in a multi-story building across several testing routes, with an average positioning error of 2.4 m. The results show that the inclusion of human activity information can reduce the overall localization error of the system and actively contribute to the better identification of floor transitions within a building. The conducted experiments demonstrated promising results and verified the effectiveness of using human activity-related information for indoor localization.
Assuntos
Corrida , Smartphone , Atividades Humanas , Humanos , CaminhadaRESUMO
Infrastructure-free Indoor Positioning Systems (IPS) are becoming popular due to their scalability and a wide range of applications. Such systems often rely on deployed Wi-Fi networks. However, their usability may be compromised, either due to scanning restrictions from recent Android versions or the proliferation of 5G technology. This raises the need for new infrastructure-free IPS independent of Wi-Fi networks. In this paper, we propose the use of magnetic field data for IPS, through Deep Neural Networks (DNN). Firstly, a dataset of human indoor trajectories was collected with different smartphones. Afterwards, a magnetic fingerprint was constructed and relevant features were extracted to train a DNN that returns a probability map of a user's location. Finally, two postprocessing methods were applied to obtain the most probable location regions. We asserted the performance of our solution against a test dataset, which produced a Success Rate of around 80%. We believe that these results are competitive for an IPS based on a single sensing source. Moreover, the magnetic field can be used as an additional information layer to increase the robustness and redundancy of current multi-source IPS.
RESUMO
The demand for easily deployable indoor localisation solutions has been growing. Although several systems have been proposed, their limitations regarding the high implementation costs hinder most of them to be widely used. Fingerprinting-based IPS (Indoor Positioning Systems) depend on characteristics pervasively available in buildings. However, such systems require indoor floor plans, which might not be available, as well as environmental fingerprints, that need to be collected through human resources intensive processes. To overcome these limitations, this paper proposes an algorithm for the automatic construction of indoor maps and fingerprints, solely depending on non-annotated crowdsourced data from smartphones. Our system relies on multiple gait-model based filtering techniques for accurate movement quantification in combination with opportunistic sensing observations. After the reconstruction of users' movement with PDR (Pedestrian Dead Reckoning) techniques, Wi-Fi measurements are clustered to partition the trajectories into segments. Similar segments, which belong to the same cluster, are identified using an adaptive approach based on a geomagnetic field distance. Finally, the floor plans are obtained through a data fusion process. Merging the acquired environmental data using the obtained floor plan, fingerprints are aligned to physical locations. Experimental results show that the proposed solution achieved comparable floor plans and fingerprints to those acquired manually, allowing the conclusion that is possible to automate the setup process of infrastructure-free IPS.
RESUMO
In Escherichia coli, under optimal conditions, protein aggregates associated with cellular aging are excluded from midcell by the nucleoid. We study the functionality of this process under sub-optimal temperatures from population and time lapse images of individual cells and aggregates and nucleoids within. We show that, as temperature decreases, aggregates become homogeneously distributed and uncorrelated with nucleoid size and location. We present evidence that this is due to increased cytoplasm viscosity, which weakens the anisotropy in aggregate displacements at the nucleoid borders that is responsible for their preference for polar localisation. Next, we show that in plasmolysed cells, which have increased cytoplasm viscosity, aggregates are also not preferentially located at the poles. Finally, we show that the inability of cells with increased viscosity to exclude aggregates from midcell results in enhanced aggregate concentration in between the nucleoids in cells close to dividing. This weakens the asymmetries in aggregate numbers between sister cells of subsequent generations required for rejuvenating cell lineages. We conclude that the process of exclusion of protein aggregates from midcell is not immune to stress conditions affecting the cytoplasm viscosity. The findings contribute to our understanding of E. coli's internal organisation and functioning, and its fragility to stressful conditions.
Assuntos
Citoplasma/química , Citoplasma/metabolismo , Proteínas de Escherichia coli/fisiologia , Escherichia coli/metabolismo , Divisão Celular , Organelas/metabolismo , Agregados Proteicos , Estresse Fisiológico , Temperatura , ViscosidadeRESUMO
In this article we propose a conceptual framework to study ensembles of conformal predictors (CP), that we call Ensemble Predictors (EP). Our approach is inspired by the application of imprecise probabilities in information fusion. Based on the proposed framework, we study, for the first time in the literature, the theoretical properties of CP ensembles in a general setting, by focusing on simple and commonly used possibilistic combination rules. We also illustrate the applicability of the proposed methods in the setting of multivariate time-series classification, showing that these methods provide better performance (in terms of both robustness, conservativeness, accuracy and running time) than both standard classification algorithms and other combination rules proposed in the literature, on a large set of benchmarks from the UCR time series archive.
RESUMO
This paper proposes a user study aimed at evaluating the impact of Class Activation Maps (CAMs) as an eXplainable AI (XAI) method in a radiological diagnostic task, the detection of thoracolumbar (TL) fractures from vertebral X-rays. In particular, we focus on two oft-neglected features of CAMs, that is granularity and coloring, in terms of what features, lower-level vs higher-level, should the maps highlight and adopting which coloring scheme, to bring better impact to the decision-making process, both in terms of diagnostic accuracy (that is effectiveness) and of user-centered dimensions, such as perceived confidence and utility (that is satisfaction), depending on case complexity, AI accuracy, and user expertise. Our findings show that lower-level features CAMs, which highlight more focused anatomical landmarks, are associated with higher diagnostic accuracy than higher-level features CAMs, particularly among experienced physicians. Moreover, despite the intuitive appeal of semantic CAMs, traditionally colored CAMs consistently yielded higher diagnostic accuracy across all groups. Our results challenge some prevalent assumptions in the XAI field and emphasize the importance of adopting an evidence-based and human-centered approach to design and evaluate AI- and XAI-assisted diagnostic tools. To this aim, the paper also proposes a hierarchy of evidence framework to help designers and practitioners choose the XAI solutions that optimize performance and satisfaction on the basis of the strongest evidence available or to focus on the gaps in the literature that need to be filled to move from opinionated and eminence-based research to one more based on empirical evidence and end-user work and preferences.
Assuntos
Processos Mentais , Radiologia , Humanos , Semântica , Coluna VertebralRESUMO
In this paper, we study human-AI collaboration protocols, a design-oriented construct aimed at establishing and evaluating how humans and AI can collaborate in cognitive tasks. We applied this construct in two user studies involving 12 specialist radiologists (the knee MRI study) and 44 ECG readers of varying expertise (the ECG study), who evaluated 240 and 20 cases, respectively, in different collaboration configurations. We confirm the utility of AI support but find that XAI can be associated with a "white-box paradox", producing a null or detrimental effect. We also find that the order of presentation matters: AI-first protocols are associated with higher diagnostic accuracy than human-first protocols, and with higher accuracy than both humans and AI alone. Our findings identify the best conditions for AI to augment human diagnostic skills, rather than trigger dysfunctional responses and cognitive biases that can undermine decision effectiveness.
Assuntos
Inteligência Artificial , HumanosRESUMO
Treatment and prevention of cardiovascular diseases often rely on Electrocardiogram (ECG) interpretation. Dependent on the physician's variability, ECG interpretation is subjective and prone to errors. Machine learning models are often developed and used to support doctors; however, their lack of interpretability stands as one of the main drawbacks of their widespread operation. This paper focuses on an Explainable Artificial Intelligence (XAI) solution to make heartbeat classification more explainable using several state-of-the-art model-agnostic methods. We introduce a high-level conceptual framework for explainable time series and propose an original method that adds temporal dependency between time samples using the time series' derivative. The results were validated in the MIT-BIH arrhythmia dataset: we performed a performance's analysis to evaluate whether the explanations fit the model's behaviour; and employed the 1-D Jaccard's index to compare the subsequences extracted from an interpretable model and the XAI methods used. Our results show that the use of the raw signal and its derivative includes temporal dependency between samples to promote classification explanation. A small but informative user study concludes this study to evaluate the potential of the visual explanations produced by our original method for being adopted in real-world clinical settings, either as diagnostic aids or training resource.