Generalization Capacity Research Articles

Application of black-box models based on artificial intelligence for the prediction of chlorine and TTHMs in the trunk network of Bogotá, Colombia

Abstract The chlorine and total trihalomethane (TTHM) concentrations are sparsely measured in the trunk network of Bogotá, Colombia, which leads to a high uncertainty level at an operational level. For this reason, this research assessed the prediction accuracy for chlorine and TTHM concentrations of two black-box models based on the following artificial intelligence techniques: artificial neural networks (ANNs) and adaptive neuro-fuzzy inference system (ANFIS) as a modelling alternative. The simulation results of a hydraulic and water quality analysis of the network in EPANET and its multi-species extension EPANET-MSX were used for training the black-box models. Subsequently, the Threat Ensemble Vulnerability Assessment-Sensor Placement Optimization Tool (TEVA-SPOT) and Evolutionary Polynomial Regression-Multi-Objective Genetic Algorithm (EPR-MOGA-XL) were jointly applied to select the most representative input variables and locations for predicting water quality at other points of the network. ANNs and ANFIS were optimized with a multi-objective approach to reach a compromise between training performance and generalization capacity. The ANFIS models had a higher mean Training and Test Nash–Sutcliffe Index (NSI) in contrast with ANNs. In general, the models had a satisfactory mean prediction performance. However, some of them did not achieve suitable Test NSI values, and the prediction accuracy for different operational statuses was limited.

A machine learning-based genetic programming approach for the sustainable production of plastic sand paver blocks

Plastic sand paver blocks (PSPB) provide a sustainable alternative by reprocessing plastic waste and decreasing reliance on environmentally hazardous materials such as concrete. They promote waste management and environmentally favorable building practices. This paper presents a novel method for estimating the compressive strength (CS) of plastic sand paver blocks based on gene expression programming (GEP) techniques. The database collected from the experimental work comprises 135 compressive strength results. Seven input parameters were involved in predicting the CS of PSPB, namely, plastic, sand, sand size, fiber percentage, fibre length, fibre diameter, and tensile strength of the fibre. Simplified mathematical expressions were used to figure out the CS. The results of GEP formulations showed that they were better in line with the experimental data, with R2 values for CS of 0.89 (training) and 0.88 (testing). The models' performance was evaluated using sensitivity analysis and statistical checks. The statistical evaluations show that the actual and predicted values are closer together, which lends credence to the GEP model's capacity to forecast PSPB CS. The sensitivity analysis showed that sand size and fibre percentage contribute more than 50% of the CS in PSPB. In addition, the results demonstrate that the proposed models are accurate and have a robust capacity for generalization and prediction. This research can improve environmental protection and economic benefit by enhancing the reuse of PSPB in producing green ecosystems.

Hybrid machine learning techniques in the management of harmful algal blooms impact

Harmful algal blooms (HABs) are episodes of high concentrations of algae that are potentially toxic for human consumption. Mollusc farming can be affected by HABs because, as filter feeders, they can accumulate high concentrations of marine biotoxins in their tissues. To avoid the risk to human consumption, harvesting is prohibited when toxicity is detected. At present, the closure of production areas is based on expert knowledge and the existence of a predictive model would help when conditions are complex and sampling is not possible. Although the concentration of toxin in meat is the method most commonly used by experts in the control of shellfish production areas, it is rarely used as a target by automatic prediction models. This is largely due to the irregularity of the data due to the established sampling programs. As an alternative, the activity status of production areas has been proposed as a target variable based on whether mollusc meat has a toxicity level below or above the legal limit. This new option is the most similar to the actual functioning of the control of shellfish production areas. For this purpose, we have made a comparison between hybrid machine learning models like Neural-Network-Adding Bootstrap (BAGNET) and Discriminative Nearest Neighbor Classification (SVM-KNN) when estimating the state of production areas. The study has been carried out in several estuaries with different levels of complexity in the episodes of algal blooms to demonstrate the generalization capacity of the models in bloom detection. As a result, we could observe that, with an average recall value of 93.41% and without dropping below 90% in any of the estuaries, BAGNET outperforms the other models both in terms of results and robustness.

Mind the Gap: Polishing Pseudo Labels for Accurate Semi-supervised Object Detection

Exploiting pseudo labels (e.g., categories and bounding boxes) of unannotated objects produced by a teacher detector have underpinned much of recent progress in semi-supervised object detection (SSOD). However, due to the limited generalization capacity of the teacher detector caused by the scarce annotations, the produced pseudo labels often deviate from ground truth, especially those with relatively low classification confidences, thus limiting the generalization performance of SSOD. To mitigate this problem, we propose a dual pseudo-label polishing framework for SSOD. Instead of directly exploiting the pseudo labels produced by the teacher detector, we take the first attempt at reducing their deviation from ground truth using dual polishing learning, where two differently structured polishing networks are elaborately developed and trained using synthesized paired pseudo labels and the corresponding ground truth for categories and bounding boxes on the given annotated objects, respectively. By doing this, both polishing networks can infer more accurate pseudo labels for unannotated objects through sufficiently exploiting their context knowledge based on the initially produced pseudo labels, and thus improve the generalization performance of SSOD. Moreover, such a scheme can be seamlessly plugged into the existing SSOD framework for joint end-to-end learning. In addition, we propose to disentangle the polished pseudo categories and bounding boxes of unannotated objects for separate category classification and bounding box regression in SSOD, which enables introducing more unannotated objects during model training and thus further improves the performance. Experiments on both PASCAL VOC and MS-COCO benchmarks demonstrate the superiority of the proposed method over existing state-of-the-art baselines. The code can be found at https://github.com/snowdusky/DualPolishLearning.

MSDC: Exploiting Multi-State Power Consumption in Non-intrusive Load Monitoring Based on a Dual-CNN Model

Non-intrusive load monitoring (NILM) aims to decompose aggregated electrical usage signal into appliance-specific power consumption and it amounts to a classical example of blind source separation tasks. Leveraging recent progress on deep learning techniques, we design a new neural NILM model {\em Multi-State Dual CNN} (MSDC). Different from previous models, MSDC explicitly extracts information about the appliance's multiple states and state transitions, which in turn regulates the prediction of signals for appliances. More specifically, we employ a dual-CNN architecture: one CNN for outputting state distributions and the other for predicting the power of each state. A new technique is invented that utilizes conditional random fields (CRF) to capture state transitions. Experiments on two real-world datasets REDD and UK-DALE demonstrate that our model significantly outperform state-of-the-art models while having good generalization capacity, achieving 6%-10% MAE gain and 33%-51% SAE gain to unseen appliances.

Boosting the performance of SOTA convolution-based networks with dimensionality reduction: An application on hyperspectral images of wine grape berries

Precision viticulture is an area that is very dependent on methods that allow for a sustainable assessment of grape maturity and, in this work, we apply two state-of-the-art (SOTA) convolution-based networks, namely InceptionTime and OmniScale 1D-CNN, to hyperspectral images of wine grape berries to estimate sugar content. Since attaining generalization capacity and processing the information in such high-dimensional data are the two biggest challenges to overcome in problems of this nature, we also study the impact of two dimensionality reduction techniques, Principal Component Analysis (PCA) and t-Distributed Stochastic Neighbor Embedding (t-SNE), on the models' performance. Both models underwent different tests with different vintages and varieties of wine grapes in the training/validation steps, as to form a true test to their generalization capacity. Our results show that both PCA and t-SNE succeed in improving the performance of these deep networks when an adequate number of components is chosen that minimizes the ratio between information loss and removing redundant features: additionally, both techniques significantly reduce computational cost, a very important trait when training deep learning models. Both models showed good generalization ability with very competitive results across different varieties and vintages even despite their significant differences in variability, which is an indicator that a relationship between spectras can be found that is reflected on sugar content values.

A general data quality evaluation framework for dynamic response monitoring of long-span bridges

The structural health monitoring system (SHM) of long-span bridges inevitably produces low-quality data. It is important to evaluate the data quality and screen out normal data. Most existing studies on data anomaly detection have focused on abnormal data with abnormal waveforms in time domain. Usually there are pseudo-normal data in the SHM system, which seems normal in time domain. However, the pseudo-normal data of bridge dynamic responses may cause wrong dynamic characteristics identification. In this study, the existing data anomaly detection was further expanded to data quality evaluation, and both obvious abnormal and pseudo-normal data are evaluated. A general data quality evaluation framework for dynamic response monitoring of long-span bridge structures was proposed. The time–frequency characteristics of monitoring data were obtained by continuous wavelet transform, and the data quality was classified and evaluated by constructing and training a convolutional neural network (CNN). The proposed framework was verified by using acceleration monitoring data of a cable-stayed bridge. The results reveal that this framework solves the problem that CNN cannot detect pseudo-normal data through time-domain characteristics, and improves the accuracy of data quality evaluation by using time–frequency information of the monitoring data. The acceleration monitoring data of a suspension bridge and a single pylon cable-stayed bridge were used to demonstrate the feasibility of the cross-object application of the proposed framework. The evaluation accuracy for the acceleration data quality of the suspension bridge exceeded 96%. After the reinforced training of a small number of pseudo-normal data samples, the evaluation accuracy for the acceleration data quality of the cables of the cable-stayed bridge reached 96.8%. The framework shows a good generalization capacity and robustness in cross-object applications.

Exploiting image translations via ensemble self-supervised learning for Unsupervised Domain Adaptation

Unsupervised Domain Adaptation (UDA) aims to improve the generalization capacity of models when they are tested on a real-world target domain by learning a model on a source labeled domain. Recently, a UDA method was proposed that addresses the adaptation problem by combining ensemble learning with self-supervised learning. However, this method uses only the source domain to pretrain the model and employs a limited amount of classifiers to create target pseudo labels. To mitigate these deficiencies, in this work, we explore the usage of image translations in combination with ensemble learning and self-supervised learning. To increase the model’s exposure to more variable pretraining data, our method creates multiple diverse image translations, which encourages the learning of domain-invariant features, desired to increase generalization. With these image translations, we are able to learn translation-specific classifiers, which also allows to maximize the amount of ensemble’s classifiers resulting in more robust target pseudo labels. In addition, we propose to use the target domain in pretraining stage to mitigate source domain bias in the network. We evaluate our method on the standard UDA benchmarks, i.e., adapting GTA V and Synthia to Cityscapes, and achieve state-of-the-art results on the mIoU metric. Extensive ablation experiments are reported to highlight the advantageous properties of our UDA strategy.

FastFusion: Deep stereo‐LiDAR fusion for real‐time high‐precision dense depth sensing

AbstractLight detection and ranging (LiDAR) and stereo cameras are two generally used solutions for perceiving 3D information. The complementary properties of these two sensor modalities motivate a fusion to derive practicable depth sensing toward real‐world applications. Promoted by deep neural network (DNN) techniques, recent works achieve superior performance on accuracy. However, the complex architecture and the sheer number of DNN parameters often lead to poor generalization capacity and non‐real‐time computing. In this paper, we present FastFusion, a three‐stage stereo‐LiDAR deep fusion scheme, which integrates the LiDAR priors into each step of classical stereo‐matching taxonomy, gaining high‐precision dense depth sensing in a real‐time manner. We integrate stereo‐LiDAR information by taking advantage of a compact binary neural network and utilize the proposed cross‐based LiDAR trust aggregation to further fuse the sparse LiDAR measurements in the back‐end of stereo matching. To align the photometrical of the input image and the depth of the estimation, we introduce a refinement network to guarantee consistency. More importantly, we present a graphic processing unit‐based acceleration framework for providing a low‐latency implementation of FastFusion, gaining both accuracy improvement and real‐time responsiveness. In the experiments, we demonstrate the effectiveness and practicability of FastFusion, which obtains a significant speedup over state‐of‐the‐art baselines while achieving comparable accuracy on depth sensing. The video demo for real‐time depth estimation of FastFusion on the real‐world driving scenario is available at https://youtu.be/nP7cls2BA8s.

Underwater acoustic target recognition based on smoothness-inducing regularization and spectrogram-based data augmentation

Underwater acoustic target recognition is a challenging task owing to the intricate underwater environments and limited data availability. Insufficient data can hinder the ability of recognition systems to support complex modeling, thus impeding their advancement. To improve the generalization capacity of recognition models, techniques such as data augmentation have been employed to simulate underwater signals and diversify data distribution. However, the complexity of underwater environments can cause the simulated signals to deviate from real scenarios, resulting in biased models that are misguided by non-true data. In this study, we propose two strategies to enhance the generalization ability of models in the case of limited data while avoiding the risk of performance degradation. First, as an alternative to traditional data augmentation, we utilize smoothness-inducing regularization, which only incorporates simulated signals in the regularization term. Additionally, we propose a specialized spectrogram-based data augmentation strategy, namely local masking and replicating (LMR), to capture inter-class relationships. Our experiments and visualization analysis demonstrate the superiority of our proposed strategies.

Generalization capacity of multi-class SVM based on Markovian resampling

The generalization performance of “All-in-one” Multi-class SVM (AIO-MSVM) based on uniformly ergodic Markovian chain (u.e.M.c.) samples is considered. We establish the fast learning rate of AIO-MSVM algorithm with u.e.M.c. samples and prove that AIO-MSVM algorithm with u.e.M.c. samples is consistent. We also propose a novel AIO-MSVM algorithm based on q-times Markovian resampling (AIO-MSVM-MR), and show the numerical investigation on the learning performance of AIO-MSVM-MR based on public datasets. The experimental studies indicate that compared to the classical AIO-MSVM algorithm and other MSVM algorithms, the proposed AIO-MSVM-MR algorithm has not only smaller misclassification rate, but also less sampling and training total time. We present some discussions on the case of unbalanced training samples, the choices of q and two technical parameters, and present some explanations on the learning performance of the proposed algorithm.

Constructing Spatio-Temporal Graphs for Face Forgery Detection

Recently, advanced development of facial manipulation techniques threatens web information security, thus, face forgery detection attracts a lot of attention. It is clear that both spatial and temporal information of facial videos contains the crucial manipulation traces, which are inevitably created during the generation process. However, most existing face forgery detectors only focus on the spatial artifacts or the temporal incoherence, and they are struggling to learn a significant and general kind of representations for manipulated facial videos. In this work, we propose to construct spatial-temporal graphs for fake videos to capture the spatial inconsistency and the temporal incoherence at the same time. To model the spatial-temporal relationship among the graph nodes, a novel forgery detector named Spatio-Temporal Graph Network (STGN) is proposed, which contains two kinds of graph-convolution-based units, the Spatial Relation Graph Unit (SRGU) and the Temporal Attention Graph Unit (TAGU). To exploit spatial information, the SRGU models the inconsistency between each pair of patches in the same frame, instead of focusing on the low-level local spatial artifacts which are vulnerable to samples created by unseen manipulation methods. And, the TAGU is proposed to model the long-distance temporal relation among the patches at the same spatial position in different frames with a graph attention mechanism based on the inter-node similarity. With the SRGU and the TAGU, our STGN can combine the discriminative power of spatial inconsistency and the generalization capacity of temporal incoherence for face forgery detection. Our STGN achieves state-of-the-art performances on several popular forgery detection datasets. Extensive experiments demonstrate both the superiority of our STGN on intra manipulation evaluation and the effectiveness for new sorts of face forgery videos on cross manipulation evaluation.

A data-driven framework for permeability prediction of natural porous rocks via microstructural characterization and pore-scale simulation.

Understanding the microstructure-property relationships of porous media is of great practical significance, based on which macroscopic physical properties can be directly derived from measurable microstructural informatics. However, establishing reliable microstructure-property mappings in an explicit manner is difficult, due to the intricacy, stochasticity, and heterogeneity of porous microstructures. In this paper, a data-driven computationalframework is presented to investigate the inherent microstructure-permeability linkage for natural porous rocks, where multiple techniques are integrated together, including microscopy imaging, stochastic reconstruction, microstructural characterization, pore-scale simulation, feature selection, and data-driven modeling. A large number of 3D digital rocks with a wide porosity range are acquired from microscopy imaging and stochastic reconstruction techniques. A broad variety of morphological descriptors are used to quantitatively characterize pore microstructures from different perspectives, and they compose the raw feature pool for feature selection. High-fidelity lattice Boltzmann simulations are conducted to resolve fluid flow passing through porous media, from which reliable permeability references are obtained. The optimal feature set that best represents permeability is identified through a performance-oriented feature selection process, upon which a cost-effective surrogate model is rapidly fitted to approximate the microstructure-permeability mapping via data-driven modeling. This surrogate model exhibits great advantages over empirical/analytical formulas in terms of prediction accuracy and generalization capacity, which can predict reliable permeability values spanning four orders of magnitude. Besides, feature selection also greatly enhances the interpretability of the data-driven prediction model, from which new insights into the mechanism of how microstructural characteristics determine intrinsic permeability are obtained.

Multi-source domain adaptation with spatio-temporal feature extractor for EEG emotion recognition

Electroencephalogram (EEG) emotion recognition suffers from cross-domain identification difficulties mainly caused by cross-subject and cross-session differences. Inspired by domain adaptation, the study constructs a novel multi-source domain adaptation with a Spatio-temporal feature extractor (MSDA-SFE) for EEG emotion recognition, which could reduce the signal differences between subjects and between sessions to make the target domain features align with the source domain features easily. In common feature learning, a Spatiotemporal feature extraction module is constructed to acquire domain-invariant characteristics of the source and target domains in which one subject is selected randomly as the target domain, and the remaining subjects are source domains. Whereafter, the obtained target domain features are paired with other source domain features respectively to generate N-1 pairs of concatenated features. And then these features are translated into N-1 domain-specific EEG features through N-1 parallel branches, where the target domain features are better aligned in the latent space, and the classification boundary of target samples is obtained by minimizing the disparity between branches. Finally, domain-specific features are fed to respective classifiers to get N-1 predictions, and their mean is put as the final result. Extensive experiments are implemented on SEED, SEED-IV, and DEAP benchmark databases. And, results indicate that the MSDA-SFE network outperforms the other comparison methods. Besides, the additional experiment also illustrates that the method has a powerful generalization capacity.

Generalizability of Machine Learning to Categorize Various Mental Illness Using Social Media Activity Patterns

Mental illness has recently become a global health issue, causing significant suffering in people’s lives and having a negative impact on productivity. In this study, we analyzed the generalization capacity of machine learning to classify various mental illnesses across multiple social media platforms (Twitter and Reddit). Language samples were gathered from Reddit and Twitter postings in discussion forums devoted to various forms of mental illness (anxiety, autism, schizophrenia, depression, bipolar disorder, and BPD). Following this process, information from 606,208 posts (Reddit) created by a total of 248,537 people and from 23,102,773 tweets was used for the analysis. We initially trained and tested machine learning models (CNN and Word2vec) using labeled Twitter datasets, and then we utilized the dataset from Reddit to assess the effectiveness of our trained models and vice versa. According to the experimental findings, the suggested method successfully classified mental illness in social media texts even when training datasets did not include keywords or when unrelated datasets were utilized for testing.

Annealing Optimization for Progressive Learning With Stochastic Approximation

In this work, we introduce a learning model designed to meet the needs of applications in which computational resources are limited, and robustness and interpretability are prioritized. Learning problems can be formulated as constrained stochastic optimization problems, with the constraints originating mainly from model assumptions that define a trade-off between complexity and performance. This trade-off is closely related to over-fitting, generalization capacity, and robustness to noise and adversarial attacks, and depends on both the structure and complexity of the model, as well as the properties of the optimization methods used. We develop an online prototype-based learning algorithm based on annealing optimization that is formulated as an online gradient-free stochastic approximation algorithm. The learning model can be viewed as an interpretable and progressively growing competitive-learning neural network model to be used for supervised, unsupervised, and reinforcement learning. The annealing nature of the algorithm contributes to minimal hyper-parameter tuning requirements, poor local minima prevention, and robustness with respect to the initial conditions. At the same time, it provides online control over the performance-complexity trade-off by progressively increasing the complexity of the learning model as needed, through an intuitive bifurcation phenomenon. Finally, the use of stochastic approximation enables the study of the convergence of the learning algorithm through mathematical tools from dynamical systems and control, and allows for its integration with reinforcement learning algorithms, constructing an adaptive state-action aggregation scheme.

Design of a Machine Learning System to Predict the Thickness of a Melanoma Lesion in a Non-Invasive Way from Dermoscopic Images.

Melanoma is the deadliest form of skin cancer, but it can be fully cured through early detection and treatment in 99% of cases. Our aim was to develop a non-invasive machine learning system that can predict the thickness of a melanoma lesion, which is a proxy for tumor progression, through dermoscopic images. This method can serve as a valuable tool in identifying urgent cases for treatment. A modern convolutional neural network architecture (EfficientNet) was used to construct a model capable of classifying dermoscopic images of melanoma lesions into three distinct categories based on thickness. We incorporated techniques to reduce the impact of an imbalanced training dataset, enhanced the generalization capacity of the model through image augmentation, and utilized five-fold cross-validation to produce more reliable metrics. Our method achieved 71% balanced accuracy for three-way classification when trained on a small public dataset of 247 melanoma images. We also presented performance projections for larger training datasets. Our model represents a new state-of-the-art method for classifying melanoma thicknesses. Performance can be further optimized by expanding training datasets and utilizing model ensembles. We have shown that earlier claims of higher performance were mistaken due to data leakage during the evaluation process.

Retinal Image Segmentation and Disease Classification using Deep Learning

Abstract: Segmenting and classifying retinal blood vessels are essential ophthalmology tasks because they reveal important details regarding the health and condition of the retina and its blood vessels. However, there is still room for improvement in terms of accuracy and robustness. Deep learning models have demonstrated significant promise in precisely and quickly segmenting and categorizing retinal pictures. In this study, we suggest the Residual U-Net (ResUNet) model for segmenting retinal arteries and the Resnet50 model for classifying retinal arteries. For enhanced feature extraction and picture segmentation, the ResUNet model, a deep learning architecture, combines the benefits of the U-Net and ResNet50 models. Accuracy was required for segmentation and classification, which gave our study's focus on the project and potential survey results in its distinctiveness. The DRIVE, CHASE-DB1, STARE, and HRF datasets, which include retinal pictures with ground truth annotations, were used to train and test our model. These data were the resource for obtaining the segmented output with analysis in accordance with the initial input supplied to the model. Another sign of the robustness of our model is its strong generalization capacity. Overall, our study shows the usefulness of the ResUNet model for retinal vascular segmentation and classification and highlights its potential for enhancing the diagnosis and treatment of retinal illnesses, which has the potential to improve further with the potential for further development.

Multi-Step Wind Power Forecasting with Stacked Temporal Convolutional Network (S-TCN)

Nowadays, wind power generation has become vital thanks to its advantages in cost, ecological friendliness, enormousness, and sustainability. However, the erratic and intermittent nature of this energy poses significant operational and management difficulties for power systems. Currently, the methods of wind power forecasting (WPF) are various and numerous. An accurate forecasting method of WPF can help system dispatchers plan unit commitment and reduce the risk of the unreliability of electricity supply. In order to improve the accuracy of short-term prediction for wind power and address the multi-step ahead forecasting, this research presents a Stacked Temporal Convolutional Network (S-TCN) model. By using dilated causal convolutions and residual connections, the suggested solution addresses the issue of long-term dependencies and performance degradation of deep convolutional models in sequence prediction. The simulation outcomes demonstrate that the S-TCN model’s training procedure is extremely stable and has a powerful capacity for generalization. Besides, the performance of the proposed model shows a higher forecasting accuracy compared to other existing neural networks like the Vanilla Long Short-Term Memory model or the Bidirectional Long Short-Term Memory model.

Generalizability of 3D CNN models for age estimation in diverse youth populations using structural MRI

Recently, several studies have investigated the neurodevelopment of psychiatric disorders using brain data acquired via structural magnetic resonance imaging (sMRI). These analyses have shown the potential of sMRI data to provide a relatively precise characterization of brain structural biomarkers. Despite these advances, a relatively unexplored question is how reliable and consistent a model is when assessing subjects from other independent datasets. In this study, we investigate the performance and generalizability of the same model architecture trained from distinct datasets comprising youths in diverse stages of neurodevelopment and with different mental health conditions. We employed models with the same 3D convolutional neural network (CNN) architecture to assess autism spectrum disorder (ASD), attention deficit hyperactivity disorder (ADHD), brain age, and a measure of dimensional psychopathology, the Child Behavior Checklist (CBCL) total score. The investigated datasets include the Autism Brain Imaging Data Exchange II (ABIDE-II, N = 580), Attention Deficit Hyperactivity Disorder (ADHD-200, N = 922), Brazilian High-Risk Cohort Study (BHRCS, N = 737), and Adolescent Brain Cognitive Development (ABCD, N = 11,031). Models’ performance and interpretability were assessed within each dataset (for diagnosis tasks) and inter-datasets (for age estimation). Despite the demographic and phenotypic differences of the subjects, all models presented significant estimations for age (p value < 0.001) within and between datasets. In addition, most models showed a moderate to high correlation in age estimation. The results, including the models' brain regions of interest (ROI), were analyzed and discussed in light of the youth neurodevelopmental structural changes. Among other interesting discoveries, we found that less confounded training datasets produce models with higher generalization capacity.