Noisy Training Research Articles

Cooperative learning of Pl@ntNet's Artificial Intelligence algorithm: How does it work and how can we improve it?

Abstract Deep learning models for plant species identification rely on large annotated datasets. The Pl@ntNet system enables global data collection by allowing users to upload and annotate plant observations, leading to noisy labels due to diverse user skills. Achieving consensus is crucial for training, but the vast scale of collected data (number of observations, users and species) makes traditional label aggregation strategies challenging. Existing methods either retain all observations, resulting in noisy training data or selectively keep those with sufficient votes, discarding valuable information. Additionally, as many species are rarely observed, user expertise cannot be evaluated as an inter‐user agreement: otherwise, botanical experts would have a lower weight in the AI training step than the average user. Our proposed label aggregation strategy aims to cooperatively train plant identification AI models. This strategy estimates user expertise as a trust score per user based on their ability to identify plant species from crowdsourced data. The trust score is recursively estimated from correctly identified species given the current estimated labels. This interpretable score exploits botanical experts' knowledge and the heterogeneity of users. Subsequently, our strategy removes unreliable observations but retains those with limited trusted annotations, unlike other approaches. We evaluate Pl@ntNet's strategy on a newly released large subset of the Pl@ntNet database focused on European flora, comprising over 6 M observations and 800 K users. This anonymized dataset of votes and observations is released openly via Lefort, Affouard, et al. (2024). We demonstrate that estimating users' skills based on the diversity of their expertise enhances labelling performance. Our findings emphasize the synergy of human annotation and data filtering in improving AI performance for a refined training dataset. We explore incorporating AI‐based votes alongside human input in the label aggregation. This can further enhance human‐AI interactions to detect unreliable observations (even with few votes).

Enhancing Autism Detection Through Gaze Analysis Using Eye Tracking Sensors and Data Attribution with Distillation in Deep Neural Networks.

Autism Spectrum Disorder (ASD) is a neurodevelopmental condition characterized by differences in social communication and repetitive behaviors, often associated with atypical visual attention patterns. In this paper, the Gaze-Based Autism Classifier (GBAC) is proposed, which is a Deep Neural Network model that leverages both data distillation and data attribution techniques to enhance ASD classification accuracy and explainability. Using data sampled by eye tracking sensors, the model identifies unique gaze behaviors linked to ASD and applies an explainability technique called TracIn for data attribution by computing self-influence scores to filter out noisy or anomalous training samples. This refinement process significantly improves both accuracy and computational efficiency, achieving a test accuracy of 94.35% while using only 77% of the dataset, showing that the proposed GBAC outperforms the same model trained on the full dataset and random sample reductions, as well as the benchmarks. Additionally, the data attribution analysis provides insights into the most influential training examples, offering a deeper understanding of how gaze patterns correlate with ASD-specific characteristics. These results underscore the potential of integrating explainable artificial intelligence into neurodevelopmental disorder diagnostics, advancing clinical research by providing deeper insights into the visual attention patterns associated with ASD.

Effects of transient, persistent, and resurgent sodium currents on excitability and spike regularity in vestibular ganglion neurons.

Vestibular afferent neurons occur as two populations with differences in spike timing regularity that are independent of rate. The more excitable regular afferents have lower current thresholds and sustained spiking responses to injected currents, while irregular afferent neurons have higher thresholds and transient responses. Differences in expression of low-voltage-activated potassium (KLV) channels are emphasized in models of spiking regularity and excitability in these neurons, leaving open the potential contributions of the voltage-gated sodium (NaV) channels responsible for the spike upstroke. We investigated the impact of different NaV current modes (transient, persistent, and resurgent) with whole-cell patch clamp experiments in mouse vestibular ganglion neurons (VGNs), the cultured and dissociated cell bodies of afferents. All VGNs had transient NaV current, many had a small persistent (non-inactivating) NaV current, and a few had resurgent current, which flows after the spike when NaV channels that were blocked are unblocked. A known NaV1.6 channel blocker decreased spike rate and altered spike waveforms in both sustained and transient VGNs and affected all three modes of NaV current. A NaV channel agonist enhanced persistent current and increased spike rate and regularity. We hypothesized that persistent and resurgent currents have different effects on sustained (regular) VGNs vs. transient (irregular) VGNs. Lacking blockers specific for the different current modes, we used modeling to isolate their effects on spiking of simulated transient and sustained VGNs, driven by simulated current steps and noisy trains of simulated EPSCs. In all simulated neurons, increasing transient NaV current increased spike rate and rate-independent regularity. In simulated sustained VGNs, adding persistent current increased both rate and rate-independent regularity, while adding resurgent current had limited impact. In transient VGNs, adding persistent current had little impact, while adding resurgent current increased both rate and rate-independent irregularity by enhancing sensitivity to synaptic noise. These experiments show that the small NaV current modes may enhance the differentiation of afferent populations, with persistent currents selectively making regular afferents more regular and resurgent currents selectively making irregular afferents more irregular.

Improving crop type mapping by integrating LSTM with temporal random masking and pixel-set spatial information

Accurate and timely crop type classification is essential for effective agricultural monitoring, cropland management, and yield estimation. Unfortunately, the complicated temporal patterns of different crops, combined with gaps and noise in satellite observations caused by clouds and rain, restrict crop classification accuracy, particularly during early seasons with limited temporal information. Although deep learning-based methods have exhibited great potential for improving crop type mapping, insufficient and noisy training data may lead them to overlook more generalizable features and derive inferior classification performance. To address these challenges, we developed a Mask Pixel-set SpatioTemporal Integration Network (Mask-PSTIN), which integrates a temporal random masking technique and a novel PSTIN model. Temporal random masking augments the training data by selectively removing certain temporal information to improve data variability, enforcing the model to learn more generalized features. The PSTIN, comprising a pixel-set aggregation encoder (PSAE) and long short-term memory (LSTM) module, effectively captures comprehensive spatiotemporal features from time-series satellite images. The effectiveness of Mask-PSTIN was evaluated across three regions with different landscapes and cropping systems. Results demonstrated that the addition of PSAE in PSTIN significantly improved crop classification accuracy compared to a basic LSTM, with average overall accuracy (OA) increasing from 80.9% to 83.9%, and the mean F1-Score (mF1) rising from 0.781 to 0.818. Incorporating temporal random masking in training led to further improvements, increasing average OA and mF1 to 87.4% and 0.865, respectively. The Mask-PSTIN significantly outperformed traditional machine learning and deep learning methods (i.e., RF, SVM, Transformer, and CNN-LSTM) in crop type mapping across all three regions. Furthermore, Mask-PSTIN enabled earlier and more accurate crop type identification before or during their developing stages compared to machine learning models. Feature importance analysis based on the gradient backpropagation algorithm revealed that Mask-PSTIN effectively leveraged multi-temporal features, exhibiting broader attention across various time steps and capturing critical crop phenological characteristics. These results suggest that Mask-PSTIN is a promising approach for improving both post-harvest and early-season crop type classification, with potential applications in agricultural management and monitoring.

Railway noise monitoring - a tool to analyse in detail the evolution of noise over time during a rolling stock renewal period

As part of the commissioning of a rail freight project, a noise monitoring system was set up to track the increase in rail freight traffic and to monitor noise over time. Noise monitoring is done by continuous measurement stations, an analysis of the classic noise exposure indexes LAeq,6-22h and LAeq,22-6h is realized, completed by an analysis of several event acoustic indexes, with a distinction between the different traffic types (Freight, regional trains). In addition, the European Regulation EU 2019/774 on the application of the technical specification for interoperability relating to the subsystem 'rolling stock - noise' to the existing freight wagons, obliges the circulation of less noisy freight trains on certain itineraries called "Quieter routes" from 8/12/2024. The projected benefit will not be limited to the nearby residents of these routes but will, by indirect consequence, enable a gradual acoustic gain on all the French rail freight routes, due to the retrofit or renewal of the existing rolling stock. The goal of the noise monitoring system is also to quantify the acoustic gains from the indirect effects of this new regulation.

Training neural networks with structured noise improves classification and generalization

Abstract The beneficial role of noise-injection in learning is a consolidated concept in the field of artificial neural networks, suggesting that even biological systems might take advantage of similar mechanisms to optimize their performance. The training-with-noise (TWN) algorithm proposed by Gardner and collaborators is an emblematic example of a noise-injection procedure in recurrent networks, which can be used to model biological neural systems. We show how adding structure to noisy training data can substantially improve the algorithm performance, allowing the network to approach perfect retrieval of the memories and wide basins of attraction, even in the scenario of maximal injected noise. We also prove that the so-called Hebbian Unlearning rule coincides with the TWN algorithm when noise is maximal and data are stable fixed points of the network dynamics.

Inverse analysis of soil hydraulic parameters of layered soil profiles using physics‐informed neural networks with unsaturated water flow models

AbstractInformation about the spatial distribution of soil hydraulic parameters is necessary for the accurate prediction of soil water flow and the coupled movement of chemicals and heat at the field scale using a process‐based model. Physics‐informed neural networks (PINNs), which can provide physical constraints in deep learning to obtain a mesh‐free solution, can be used to inversely estimate soil hydraulic parameters from less and noisy training data. Previous studies using PINNs have successfully estimated soil hydraulic parameters for homogeneous soil but estimating such parameters of layered soil profiles where the interface depth and the parameters are unknown still has some difficulties. The objective of this study was to develop PINNs to inversely estimate the distribution of soil hydraulic parameters, such as saturated hydraulic conductivity and α and n of the Mualem–van Genuchten model directly within layered soil profiles by predicting changes in the pressure head from training data based on simulation results at given depths during infiltration. The impact of factors affecting PINNs performance, such as the weights assigned to each component of the loss function, time range used in error computations, and number of samples used to assess the physical constraint, was investigated. By assigning a larger weight to the physical constraint and excluding the earlier stage of infiltration in the loss function, the changes in the pressure head and the three soil hydraulic parameter distributions within the layered soil were successfully estimated. The developed PINNs can be further applied to more complex soils and can be improved.

An uncertainty perception metric network for machinery fault diagnosis under limited noisy source domain and scarce noisy unknown domain

Deep learning has made notable advances in intelligent fault diagnosis. However, industrial application of deep learning models faces challenges due to noise interference and scarce labeled samples. Targeting the above problems, this paper proposes a metric network-based diagnostic method. For the problem of noise, a multi-scale cross feature extraction module (MSCM) is constructed to mine key classification information under noise interference to improve fault identifiability. Different from the current approach to metric learning, this paper models the uncertainty of the similarity between ‘query sample-class prototype’, and develops corresponding loss function for more effective perception, thereby better improving the fault recognition ability of the model under limited noisy source domain and scarce noisy unknown domain. Meanwhile, to visualize the decision-making process of the model under uncertainty and improve interpretability, this paper develops a novel colony-based class activation mapping (Colony-CAM) tool, which is more reliable and focused. The proposed method is compared with five baselines across three datasets. It achieved leading diagnostic accuracies of 98.55% and 94.33% with 70 and 40 noisy training samples, respectively.

Mask the Unknown: Assessing Different Strategies to Handle Weak Annotations in the MICCAI2023 Mediastinal Lymph Node Quantification Challenge

Pathological lymph node delineation is crucial in cancer diagnosis, progression assessment, and treatment planning. The MICCAI 2023 Lymph Node Quantification Challenge published the first public dataset for pathological lymph node segmentation in the mediastinum. As lymph node annotations are expensive, the challenge was formed as a weakly supervised learning task, where only a subset of all lymph nodes in the training set have been annotated. For the challenge submission, multiple methods for training on these weakly supervised data were explored, including noisy label training, loss masking of unlabeled data, and an approach that integrated the TotalSegmentator toolbox as a form of pseudo labeling in order to reduce the number of unknown voxels. Furthermore, multiple public TCIA datasets were incorporated into the training to improve the performance of the deep learning model. Our submitted model achieved a Dice score of 0.628 and an average symmetric surface distance of 5.8~mm on the challenge test set. With our submitted model, we accomplished the third rank in the MICCAI2023 LNQ challenge. A finding of our analysis was that the integration of all visible, including non-pathological lymph nodes improved the overall segmentation performance on pathological lymph nodes of the test set. Furthermore, segmentation models trained only on clinically enlarged lymph nodes, as given in the challenge scenario, could not generalize to smaller pathological lymph nodes. The code and model for the challenge submission are available at <a href='https://gitlab.lrz.de/compai/MediastinalLymphNodeSegmentation'>https://gitlab.lrz.de/compai/MediastinalLymphNodeSegmentation</a>

Identifying Plausible Labels from Noisy Training Data for a Land Use and Land Cover Classification Application in Amazônia Legal

Most studies in the field of land use and land cover (LULC) classification in remote sensing rely on supervised classification, which requires a substantial amount of accurate label data. However, reliable data are often not immediately available, and are obtained through time-consuming manual labor. One potential solution to this problem is the use of already available classification maps, which may not be the true ground truth and may contain noise from multiple possible sources. This is also true for the classification maps of the MapBiomas project, which provides land use and land cover (LULC) maps on a yearly basis, classifying the Amazon basin into more than 24 classes based on the Landsat data. In this study, we utilize the Sentinel-2 data with a higher spatial resolution in conjunction with the MapBiomas maps to evaluate a proposed noise removal method and to improve classification results. We introduce a novel noise detection method that relies on identifying anchor points in feature space through clustering with self-organizing maps (SOM). The pixel label is relabeled using nearest neighbor rules, or can be removed if it is unknown. A challenge in this approach is the quantification of noise in such a real-world dataset. To overcome this problem, highly reliable validation sets were manually created for quantitative performance assessment. The results demonstrate a significant increase in overall accuracy compared to MapBiomas labels, from 79.85% to 89.65%. Additionally, we trained the L2HNet using both MapBiomas labels and the filtered labels from our approach. The overall accuracy for this model reached 93.75% with the filtered labels, compared to the baseline of 74.31%. This highlights the significance of noise detection and filtering in remote sensing, and emphasizes the need for further research in this area.

Multitask Learning Strategy with Pseudo-Labeling: Face Recognition, Facial Landmark Detection, and Head Pose Estimation

Most facial analysis methods perform well in standardized testing but not in real-world testing. The main reason is that training models cannot easily learn various human features and background noise, especially for facial landmark detection and head pose estimation tasks with limited and noisy training datasets. To alleviate the gap between standardized and real-world testing, we propose a pseudo-labeling technique using a face recognition dataset consisting of various people and background noise. The use of our pseudo-labeled training dataset can help to overcome the lack of diversity among the people in the dataset. Our integrated framework is constructed using complementary multitask learning methods to extract robust features for each task. Furthermore, introducing pseudo-labeling and multitask learning improves the face recognition performance by enabling the learning of pose-invariant features. Our method achieves state-of-the-art (SOTA) or near-SOTA performance on the AFLW2000-3D and BIWI datasets for facial landmark detection and head pose estimation, with competitive face verification performance on the IJB-C test dataset for face recognition. We demonstrate this through a novel testing methodology that categorizes cases as soft, medium, and hard based on the pose values of IJB-C. The proposed method achieves stable performance even when the dataset lacks diverse face identifications.

Fake news detection in low-resource languages: A novel hybrid summarization approach

The proliferation of fake news across languages and domains on social media platforms poses a significant societal threat. Current automatic detection methods for low-resource languages (e.g., Swahili, Indonesian and other low-resource languages) face limitations due to two factors: sequential length restrictions in pre-trained language models (PLMs) like multilingual bidirectional encoder representation from transformers (mBERT), and the presence of noisy training data. This work proposes a novel and efficient multilingual fake news detection (MFND) approach that addresses these challenges. Our solution leverages a hybrid extractive and abstractive summarization strategy to extract only the most relevant content from news articles. This significantly reduces data length while preserving crucial information for fake news classification. The pre-processed data is then fed into mBERT for classification. Extensive evaluations on a publicly available multilingual dataset demonstrate the superiority of our approach compared to state-of-the-art (SOTA) methods. Our analysis, both quantitative and qualitative, highlights the strengths of this method, achieving new performance benchmarks and emphasizing the impact of content condensation on model accuracy and efficiency. This framework paves the way for faster, more accurate MFND, fostering more robust information ecosystems.

Video Moment Retrieval With Noisy Labels.

Video moment retrieval (VMR) aims to localize the target moment in an untrimmed video according to the given nature language query. The existing algorithms typically rely on clean annotations to train their models. However, making annotations by human labors may introduce much noise. Thus, the video moment retrieval models will not be well trained in practice. In this article, we present a simple yet effective video moment retrieval framework via bottom-up schema, which is in end-to-end manners and robust to noisy label training. Specifically, we extract the multimodal features by syntactic graph convolutional networks and multihead attention layers, which are fused by the cross gates and the bilinear approach. Then, the feature pyramid networks are constructed to encode plentiful scene relationships and capture high semantics. Furthermore, to mitigate the effects of noisy annotations, we devise the multilevel losses characterized by two levels: a frame-level loss that improves noise tolerance and an instance-level loss that reduces adverse effects of negative instances. For the frame level, we adopt the Gaussian smoothing to regard noisy labels as soft labels through the partial fitting. For the instance level, we exploit a pair of structurally identical models to let them teach each other during iterations. This leads to our proposed robust video moment retrieval model, which experimentally and significantly outperforms the state-of-the-art approaches on standard public datasets ActivityCaption and textually annotated cooking scene (TACoS). We also evaluate the proposed approach on the different manual annotation noises to further demonstrate the effectiveness of our model.

Knockoffs-SPR: Clean Sample Selection in Learning With Noisy Labels.

A noisy training set usually leads to the degradation of the generalization and robustness of neural networks. In this article, we propose a novel theoretically guaranteed clean sample selection framework for learning with noisy labels. Specifically, we first present a Scalable Penalized Regression (SPR) method, to model the linear relation between network features and one-hot labels. In SPR, the clean data are identified by the zero mean-shift parameters solved in the regression model. We theoretically show that SPR can recover clean data under some conditions. Under general scenarios, the conditions may be no longer satisfied; and some noisy data are falsely selected as clean data. To solve this problem, we propose a data-adaptive method for Scalable Penalized Regression with Knockoff filters (Knockoffs-SPR), which is provable to control the False-Selection-Rate (FSR) in the selected clean data. To improve the efficiency, we further present a split algorithm that divides the whole training set into small pieces that can be solved in parallel to make the framework scalable to large datasets. While Knockoffs-SPR can be regarded as a sample selection module for a standard supervised training pipeline, we further combine it with a semi-supervised algorithm to exploit the support of noisy data as unlabeled data. Experimental results on several benchmark datasets and real-world noisy datasets show the effectiveness of our framework and validate the theoretical results of Knockoffs-SPR.

Mini-Workshop: Interpolation and Over-parameterization in Statistics and Machine Learning​

In recent years it has become clear that, contrary to traditional statistical beliefs, methods that interpolate (fit exactly) the noisy training data, can still be statistically optimal. In particular, this phenomenon of “be- nign overfitting” or “harmless interpolation” seems to be close to the practical regimes of modern deep learning systems, and, arguably, underlies many of their behaviors. This workshop brought together experts on the emerging theory of interpolation in statistical methods, its theoretical foundations and applications to machine learning and deep learning.

Enhancing Colorectal Cancer Tumor Bud Detection Using Deep Learning from Routine H&E-Stained Slides.

Tumor budding refers to a cluster of one to four tumor cells located at the tumor-invasive front. While tumor budding is a prognostic factor for colorectal cancer, counting and grading tumor budding are time consuming and not highly reproducible. There could be high inter- and intra-reader disagreement on H&E evaluation. This leads to the noisy training (imperfect ground truth) of deep learning algorithms, resulting in high variability and losing their ability to generalize on unseen datasets. Pan-cytokeratin staining is one of the potential solutions to enhance the agreement, but it is not routinely used to identify tumor buds and can lead to false positives. Therefore, we aim to develop a weakly-supervised deep learning method for tumor bud detection from routine H&E-stained images that does not require strict tissue-level annotations. We also propose Bayesian Multiple Instance Learning (BMIL) that combines multiple annotated regions during the training process to further enhance the generalizability and stability in tumor bud detection. Our dataset consists of 29 colorectal cancer H&E-stained images that contain 115 tumor buds per slide on average. In six-fold cross-validation, our method demonstrated an average precision and recall of 0.94, and 0.86 respectively. These results provide preliminary evidence of the feasibility of our approach in improving the generalizability in tumor budding detection using H&E images while avoiding the need for non-routine immunohistochemical staining methods.

Spatial-Logic-Aware Weakly Supervised Learning for Flood Mapping on Earth Imagery

Flood mapping on Earth imagery is crucial for disaster management, but its efficacy is hampered by the lack of high-quality training labels. Given high-resolution Earth imagery with coarse and noisy training labels, a base deep neural network model, and a spatial knowledge base with label constraints, our problem is to infer the true high-resolution labels while training neural network parameters. Traditional methods are largely based on specific physical properties and thus fall short of capturing the rich domain constraints expressed by symbolic logic. Neural-symbolic models can capture rich domain knowledge, but existing methods do not address the unique spatial challenges inherent in flood mapping on high-resolution imagery. To fill this gap, we propose a spatial-logic-aware weakly supervised learning framework. Our framework integrates symbolic spatial logic inference into probabilistic learning in a weakly supervised setting. To reduce the time costs of logic inference on vast high-resolution pixels, we propose a multi-resolution spatial reasoning algorithm to infer true labels while training neural network parameters. Evaluations of real-world flood datasets show that our model outperforms several baselines in prediction accuracy. The code is available at https://github.com/spatialdatasciencegroup/SLWSL.

Occlusion-Aware Self-Supervised Monocular 6D Object Pose Estimation.

6D object pose estimation is a fundamental yet challenging problem in computer vision. Convolutional Neural Networks (CNNs) have recently proven to be capable of predicting reliable 6D pose estimates even under monocular settings. Nonetheless, CNNs are identified as being extremely data-driven, and acquiring adequate annotations is oftentimes very time-consuming and labor intensive. To overcome this limitation, we propose a novel monocular 6D pose estimation approach by means of self-supervised learning, removing the need for real annotations. After training our proposed network fully supervised with synthetic RGB data, we leverage current trends in noisy student training and differentiable rendering to further self-supervise the model on these unsupervised real RGB(-D) samples, seeking for a visually and geometrically optimal alignment. Moreover, employing both visible and amodal mask information, our self-supervision becomes very robust towards challenging scenarios such as occlusion. Extensive evaluations demonstrate that our proposed self-supervision outperforms all other methods relying on synthetic data or employing elaborate techniques from the domain adaptation realm. Noteworthy, our self-supervised approach consistently improves over its synthetically trained baseline and often almost closes the gap towards its fully supervised counterpart.

Wavelet-Improved Score-Based Generative Model for Medical Imaging.

The score-based generative model (SGM) has demonstrated remarkable performance in addressing challenging under-determined inverse problems in medical imaging. However, acquiring high-quality training datasets for these models remains a formidable task, especially in medical image reconstructions. Prevalent noise perturbations or artifacts in low-dose Computed Tomography (CT) or under-sampled Magnetic Resonance Imaging (MRI) hinder the accurate estimation of data distribution gradients, thereby compromising the overall performance of SGMs when trained with these data. To alleviate this issue, we propose a wavelet-improved denoising technique to cooperate with the SGMs, ensuring effective and stable training. Specifically, the proposed method integrates a wavelet sub-network and the standard SGM sub-network into a unified framework, effectively alleviating inaccurate distribution of the data distribution gradient and enhancing the overall stability. The mutual feedback mechanism between the wavelet sub-network and the SGM sub-network empowers the neural network to learn accurate scores even when handling noisy samples. This combination results in a framework that exhibits superior stability during the learning process, leading to the generation of more precise and reliable reconstructed images. During the reconstruction process, we further enhance the robustness and quality of the reconstructed images by incorporating regularization constraint. Our experiments, which encompass various scenarios of low-dose and sparse-view CT, as well as MRI with varying under-sampling rates and masks, demonstrate the effectiveness of the proposed method by significantly enhanced the quality of the reconstructed images. Especially, our method with noisy training samples achieves comparable results to those obtained using clean data. Our code at https://zenodo.org/record/8266123.

Uncertainty Quantification of a Machine Learning Model for Identification of Isolated Nonlinearities with Conformal Prediction

Abstract Structural nonlinearities are often spatially localized, such joints and interfaces, localized damage, or isolated connections, in an otherwise linearly behaving system. Quinn and Brink [12] modeled this localized nonlinearity as a deviatoric force component. In other previous work [13], the authors proposed a physics-informed machine learning framework to determine the deviatoric force from measurements obtained only at the boundary of the nonlinear region, assuming a noise-free environment. However, in real experimental applications, the data are expected to contain noise from a variety of sources. In the present work, we explore the sensitivity of the trained network by comparing the network responses when trained on deterministic (“noise-free”) model data and model data with additive noise (“noisy”). As the neural network does not yield a closed-form transformation from the input distribution to the response distribution, we leverage the use of conformal sets to build an illustration of sensitivity. Through the conformal set assumption of exchangeability, we may build a distribution-free prediction interval for both network responses of the clean and noisy training sets. This work will explore the application of conformal sets for uncertainty quantification of a deterministic structure-preserving neural network and its deployment in a structural health monitoring framework to detect deviations from a baseline state based on noisy measurements.