Multi-task Machine Learning Research Articles

Accurate and Efficient Structure Elucidation from Routine One-Dimensional NMR Spectra Using Multitask Machine Learning

Accurate and Efficient Structure Elucidation from Routine One-Dimensional NMR Spectra Using Multitask Machine Learning

Multitask machine learning-based tumor-associated collagen signatures predict peritoneal recurrence and disease-free survival in gastric cancer.

Accurate prediction of peritoneal recurrence for gastric cancer (GC) is crucial in clinic. The collagen alterations in tumor microenvironment affect the migration and treatment response of cancer cells. Herein, we proposed multitask machine learning-based tumor-associated collagen signatures (TACS), which are composed of quantitative collagen features derived from multiphoton imaging, to simultaneously predict peritoneal recurrence (TACSPR) and disease-free survival (TACSDFS). Among 713 consecutive patients, with 275 in training cohort, 222 patients in internal validation cohort, and 216 patients in external validation cohort, we developed and validated a multitask machine learning model for simultaneously predicting peritoneal recurrence (TACSPR) and disease-free survival (TACSDFS). The accuracy of the model for prediction of peritoneal recurrence and prognosis as well as its association with adjuvant chemotherapy were evaluated. The TACSPR and TACSDFS were independently associated with peritoneal recurrence and disease-free survival in three cohorts, respectively (all P < 0.001). The TACSPR demonstrated a favorable performance for peritoneal recurrence in all three cohorts. In addition, the TACSDFS also showed a satisfactory accuracy for disease-free survival among included patients. For stage II and III diseases, adjuvant chemotherapy improved the survival of patients with low TACSPR and low TACSDFS, or high TACSPR and low TACSDFS, or low TACSPR and high TACSDFS, but had no impact on patients with high TACSPR and high TACSDFS. The multitask machine learning model allows accurate prediction of peritoneal recurrence and survival for GC and could distinguish patients who might benefit from adjuvant chemotherapy.

Damage detection and localization in sealed spent nuclear fuel dry storage canisters using multi-task machine learning classifiers

Spent nuclear fuel (SNF) assemblies (FAs), composed of bundled radioactive fuel rods, are stored in stainless-steel canisters as an interim dry storage option until permanent storage solutions are available. Accidental damage to these canisters may occur during handling or transportation events. If such events occur, it is necessary to assess the integrity of FAs before long-term storage. Since the canisters are sealed shut and can only be opened in special handling facilities, non-destructive evaluation (NDE) is critical for detecting the FAs from the canister's exterior surface. In this study, two multi-task machine learning (ML) classifiers, a k-nearest neighbors (k-NN) and a convolutional neural network (CNN), are developed to simultaneously detect and localize internal FA damage. The classifiers were trained and tested on a dataset collected via experimental modal analysis on a 2/3-scale mock-up canister. The canister was excited from the bottom plate while accelerometers were also attached on the bottom plate at arbitrary locations to record the structural response. The differences in frequency response functions (FRFs) between an intact fully loaded canister basket (FLCB) system and the canister with simulated internal damage were calculated and used as input to the ML models. Results showed that both classifiers achieved high accuracy on the testing set. The k-NN classifier produced macro-F1 scores of 1.00 for the damage detection task and 0.996 for the localization task. The macro-F1 scores of the CNN were 0.991 for damage detection and 0.964 for damage localization. Additionally, dropout layers were added to the fully connected layers of the CNN to introduce model uncertainty. By testing the model 1,000 times, probability density functions (PDFs) were generated, and it was confirmed that the CNN produced confident predictions in both the damage detection and localization tasks. This multi-task ML method contributes to advancing NDE of SNF canisters and holds potential for field applications in inspecting actual SNF canisters.

Gas permeability, diffusivity, and solubility in polymers: Simulation-experiment data fusion and multi-task machine learning

Machine learning (ML) models for predicting gas permeability through polymers have traditionally relied on experimental data. While these models exhibit robustness within familiar chemical domains, reliability wanes when applied to new spaces. To address this challenge, we present a multi-tiered multi-task learning framework empowered with advanced machine-crafted polymer fingerprinting algorithms and data fusion techniques. This framework combines scarce “high-fidelity” experimental data with abundant diverse “low-fidelity” simulation or synthetic data, resulting in predictive models that display a high level of generalizability across novel chemical spaces. Additionally, this multi-task scheme capitalizes on known physics and interrelated properties, such as gas diffusivity and solubility, both of which are closely tied to permeability. By amalgamating high throughput generated simulation data with available experimental data for gas permeability, diffusivity, and solubility for various gases, we construct multi-task deep learning models. These models can simultaneously predict all three properties for all gases under consideration, with markedly enhanced predictive accuracy, particularly compared to traditional models reliant solely on experimental data for a singular property. This strategy underscores the potential of coupling high-throughput classical simulations with data fusion methodologies to yield state-of-the-art property predictors, especially when experimental data for targeted properties is scarce.

OmniJet-α: the first cross-task foundation model for particle physics

Abstract Foundation models are multi-dataset and multi-task machine learning methods that once pre-trained can be fine-tuned for a large variety of downstream applications. The successful development of such general-purpose models for physics data would be a major breakthrough as they could improve the achievable physics performance while at the same time drastically reduce the required amount of training time and data. We report significant progress on this challenge on several fronts. First, a comprehensive set of evaluation methods is introduced to judge the quality of an encoding from physics data into a representation suitable for the autoregressive generation of particle jets with transformer architectures (the common backbone of foundation models). These measures motivate the choice of a higher-fidelity tokenization compared to previous works. Finally, we demonstrate transfer learning between an unsupervised problem (jet generation) and a classic supervised task (jet tagging) with our new OmniJet-α model. This is the first successful transfer between two different and actively studied classes of tasks and constitutes a major step in the building of foundation models for particle physics.

Machine-learning assisted analysis on coupled fluid-dynamics and electrochemical processes in interdigitated channel for iron-chromium flow batteries

This study explores the impact of interdigitated flow channel spacing on electrolyte distribution and flow velocity in porous electrodes, thereby affecting pump consumption and system efficiency. Utilizing a 3D electrochemical flow-coupled model, the research investigates the electrochemical performance and energy efficiency across various channel spacings, specific flow rates, and current densities. It elucidates the mechanisms by which interdigitated channel spacing influences battery performance. Through controlled trial-and-error, the optimal spacing for flow channels in Iron-Chromium Redox Flow Batteries (ICRFBs) was determined to be 4 mm. At a current density of 140 mA/cm2, the voltage efficiency reached 86.3 %, and the pump-based voltage efficiency achieved 85.9 %. To quickly and efficiently explore the impact of each individual parameter on battery efficiency and identify the optimal operating conditions, this study further developed a multi-task machine learning (ML) model. Initially trained on simulation data, the model achieved high predictive accuracy (R2 > 0.88) and effectively linked the interdigitated flow field design of ICRFBs with current density, specific flow rate, and channel spacing. The ML model was then validated through further investigation to ensure its accuracy and to achieve optimum performance efficiency as recommended by the model. This integrated approach offers a comprehensive understanding of RFBs performance under varying conditions, driving advancements in RFBs technology.

Multi-Task ADME/PK prediction at industrial scale: leveraging large and diverse experimental datasets.

ADME (Absorption, Distribution, Metabolism, Excretion) properties are key parameters to judge whether a drug candidate exhibits a desired pharmacokinetic (PK) profile. In this study, we tested multi-task machine learning (ML) models to predict ADME and animal PK endpoints trained on in-house data generated at Boehringer Ingelheim. Models were evaluated both at the design stage of a compound (i. e., no experimental data of test compounds available) and at testing stage when a particular assay would be conducted (i. e., experimental data of earlier conducted assays may be available). Using realistic time-splits, we found a clear benefit in performance of multi-task graph-based neural network models over single-task model, which was even stronger when experimental data of earlier assays is available. In an attempt to explain the success of multi-task models, we found that especially endpoints with the largest numbers of data points (physicochemical endpoints, clearance in microsomes) are responsible for increased predictivity in more complex ADME and PK endpoints. In summary, our study provides insight into how data for multiple ADME/PK endpoints in a pharmaceutical company can be best leveraged to optimize predictivity of ML models.

Accelerating Molecular Vibrational Spectra Simulations with a Physically Informed Deep Learning Model.

In recent years, machine learning (ML) surrogate models have emerged as an indispensable tool to accelerate simulations of physical and chemical processes. However, there is still a lack of ML models that can accurately predict molecular vibrational spectra. Here, we present a highly efficient multitask ML surrogate model termed Vibrational Spectra Neural Network (VSpecNN), to accurately calculate infrared (IR) and Raman spectra based on dipole moments and polarizabilities obtained on-the-fly via ML-enhanced molecular dynamics simulations. The methodology is applied to pyrazine, a prototypical polyatomic chromophore. The VSpecNN-predicted energies are well within the chemical accuracy (1 kcal/mol), and the errors for VSpecNN-predicted forces are only half of those obtained from a popular high-performance ML model. Compared to the ab initio reference, the VSpecNN-predicted frequencies of IR and Raman spectra differ only by less than 5.87 cm-1, and the intensities of IR spectra and the depolarization ratios of Raman spectra are well reproduced. The VSpecNN model developed in this work highlights the importance of constructing highly accurate neural network potentials for predicting molecular vibrational spectra.

A novel multi-task machine learning classifier for rare disease patterning using cardiac strain imaging data

To provide accurate predictions, current machine learning-based solutions require large, manually labeled training datasets. We implement persistent homology (PH), a topological tool for studying the pattern of data, to analyze echocardiography-based strain data and differentiate between rare diseases like constrictive pericarditis (CP) and restrictive cardiomyopathy (RCM). Patient population (retrospectively registered) included those presenting with heart failure due to CP (n = 51), RCM (n = 47), and patients without heart failure symptoms (n = 53). Longitudinal, radial, and circumferential strains/strain rates for left ventricular segments were processed into topological feature vectors using Machine learning PH workflow. In differentiating CP and RCM, the PH workflow model had a ROC AUC of 0.94 (Sensitivity = 92%, Specificity = 81%), compared with the GLS model AUC of 0.69 (Sensitivity = 65%, Specificity = 66%). In differentiating between all three conditions, the PH workflow model had an AUC of 0.83 (Sensitivity = 68%, Specificity = 84%), compared with the GLS model AUC of 0.68 (Sensitivity = 52% and Specificity = 76%). By employing persistent homology to differentiate the “pattern” of cardiac deformations, our machine-learning approach provides reasonable accuracy when evaluating small datasets and aids in understanding and visualizing patterns of cardiac imaging data in clinically challenging disease states.

Seismic Characterization of Subsurface Structures at Rock Valley, Nevada

ABSTRACT The Source Physics Experiment (SPE) aims at improving explosion monitoring techniques by investigating source characteristics of chemical explosions in geologic formations. One of the critical tasks in Rock Valley Direct Comparison (RV/DC), SPE phase III, is to prepare for the main experiment by characterizing the subsurface structures at the test site. Based on the seismic data acquired during an accelerated-weight-drop (AWD) seismic survey at Rock Valley, we first pick the P-wave first-arrival travel times and derive a P-wave velocity model using the adjoint-state first-arrival travel-time tomography. We then apply reverse-time migration to the processed seismic data and obtain high-resolution images of the subsurface structures along the two main survey lines. Our migration results show several reflectors corresponding to major geologic formation boundaries. We employ a multitask machine learning model to enhance the reverse-time migration images and identify faults from these images. We find that our automatically picked faults correlate well with the locations of known faults in the region in addition to many geologically undetected faults. Our subsurface characterization results refine our understanding of the geology in this region and provide valuable velocity and structural information for RV/DC geologic model building and fault identification.

Analyzing the Effects of Combining Gradient Conflict Mitigation Methods in Multi-Task Learning

Multi-task machine learning approaches involve training a single model on multiple tasks at once to increase performance and efficiency over multiple singletask models trained individually on each task. When such a multi-task model is trained to perform multiple unrelated tasks, performance can degrade significantly since unrelated tasks often have gradients that vary widely in direction. These conflicting gradients may destructively interfere with each other, causing weights learned during the training of some tasks to become unlearned during the training of others. The research selects three existing methods to mitigate this problem: Project Conflicting Gradients (PCGrad), Modulation Module, and Language-Specific Subnetworks (LaSS). It explores how the application of different combinations of these methods affects the performance of a convolutional neural network on a multi-task image classification problem. The image classification problem used as a benchmark utilizes a dataset of 4,503 leaf images to create two separate tasks: the classification of plants and the detection of disease from leaf images. Experiment results on this problem show performance benefits over singular mitigation methods, with a combination of PCGrad and LaSS obtaining a task-averaged F1 score of 0.84686. This combination outperforms individual mitigation approaches by 0.01870, 0.02682, and 0.02434 for PCGrad, Modulation Module, and LaSS, respectively in terms of F1 score.

Multi-Task Learning for Design Under Uncertainty With Multi-Fidelity Partially Observed Information

Abstract The assessment of system performance and identification of failure mechanisms in complex engineering systems often requires the use of computation-intensive finite element software or physical experiments, which are both costly and time-consuming. Moreover, when accounting for uncertainties in the manufacturing process, material properties, and loading conditions, the process of reliability-based design optimization (RBDO) for complex engineering systems necessitates the repeated execution of expensive tasks throughout the optimization process. To address this problem, this paper proposes a novel methodology for RBDO. First, a multi-fidelity surrogate modeling strategy is presented, leveraging partially observed information (POI) from diverse sources with varying fidelity and dimensionality to reduce computational cost associated with evaluating expensive high-dimensional complex systems. Second, a multi-task surrogate modeling framework is proposed to address the concurrent evaluation of multiple constraints for each design point. The multi-task framework aids in the development of surrogate models and enhances the effectiveness of reliability analysis and design optimization. The proposed multi-fidelity multi-task machine learning model utilizes a Bayesian framework, which significantly improves the performance of the predictive model and provides uncertainty quantification of the prediction. Additionally, the model provides a highly accurate and efficient framework for reliability-based design optimization through knowledge sharing. The proposed method was applied to two design case studies. By incorporating POI from various sources, the proposed approach improves the accuracy and efficiency of system performance prediction, while simultaneously addressing the cost and complexity associated with the design of complex systems.

Multitask learning to predict successful weaning in critically ill ventilated patients: A retrospective analysis of the MIMIC-IV database.

Weaning is an essential issue in critical care. This study explores the efficacy of multitask learning models in predicting successful weaning in critically ill ventilated patients using the Medical Information Mart for Intensive Care (MIMIC) IV database. We employed a multitask learning framework with a shared bottom network to facilitate common knowledge extraction across all tasks. We used the Shapley additive explanations (SHAP) plot and partial dependence plot (PDP) for model explainability. Furthermore, we conducted an error analysis to assess the strength and limitation of the model. Area under receiver operating characteristic curve (AUROC), calibration plot and decision curve analysis were used to determine the performance of the model. A total of 7758 critically ill patients were included in the analyses, and 78.5% of them were successfully weaned. Multitask learning combined with spontaneous breath trial achieved a higher performance to predict successful weaning compared with multitask learning combined with shock and mortality (area under receiver operating characteristic curve, AUROC, 0.820 ± 0.002 vs 0.817 ± 0.001, p < 0.001). We assessed the performance of the model using calibration and decision curve analyses and further interpreted the model through SHAP and PDP plots. The error analysis identified a relatively high error rate among those with low disease severities, including low mean airway pressure and high enteral feeding. We demonstrated that multitask machine learning increased predictive accuracy for successful weaning through combining tasks with a high inter-task relationship. The model explainability and error analysis should enhance trust in the model.

Enhancing Knowledge Reusability: A Distributed Multitask Machine Learning Approach

Enhancing Knowledge Reusability: A Distributed Multitask Machine Learning Approach

Analyzing Surgical Technique in Diverse Open Surgical Videos With Multitask Machine Learning

To overcome limitations of open surgery artificial intelligence (AI) models by curating the largest collection of annotated videos and to leverage this AI-ready data set to develop a generalizable multitask AI model capable of real-time understanding of clinically significant surgical behaviors in prospectively collected real-world surgical videos. The study team programmatically queried open surgery procedures on YouTube and manually annotated selected videos to create the AI-ready data set used to train a multitask AI model for 2 proof-of-concept studies, one generating surgical signatures that define the patterns of a given procedure and the other identifying kinematics of hand motion that correlate with surgeon skill level and experience. The Annotated Videos of Open Surgery (AVOS) data set includes 1997 videos from 23 open-surgical procedure types uploaded to YouTube from 50 countries over the last 15 years. Prospectively recorded surgical videos were collected from a single tertiary care academic medical center. Deidentified videos were recorded of surgeons performing open surgical procedures and analyzed for correlation with surgical training. The multitask AI model was trained on the AI-ready video data set and then retrospectively applied to the prospectively collected video data set. Analysis of open surgical videos in near real-time, performance on AI-ready and prospectively collected videos, and quantification of surgeon skill. Using the AI-ready data set, the study team developed a multitask AI model capable of real-time understanding of surgical behaviors-the building blocks of procedural flow and surgeon skill-across space and time. Through principal component analysis, a single compound skill feature was identified, composed of a linear combination of kinematic hand attributes. This feature was a significant discriminator between experienced surgeons and surgical trainees across 101 prospectively collected surgical videos of 14 operators. For each unit increase in the compound feature value, the odds of the operator being an experienced surgeon were 3.6 times higher (95% CI, 1.67-7.62; P = .001). In this observational study, the AVOS-trained model was applied to analyze prospectively collected open surgical videos and identify kinematic descriptors of surgical skill related to efficiency of hand motion. The ability to provide AI-deduced insights into surgical structure and skill is valuable in optimizing surgical skill acquisition and ultimately improving surgical care.

Multi-task machine learning models for simultaneous prediction of tissue-to-blood partition coefficients of chemicals in mammals

Tissue-to-blood partition coefficients (Ptb) are crucial for assessing the distribution of chemicals in organisms. Given the lack of experimental data and laborious nature of experimental methods, there is an urgent need to develop efficient predictive models. With the help of machine learning algorithms, i,e., random forest (RF), and artificial neural network (ANN), this study developed multi-task (MT) models that can simultaneously predict Ptb values for various mammalian tissues, including liver, muscle, brain, lung, and adipose. Single-task (ST) models using partial least squares regression, RF, and ANN algorithms for each endpoint were established for comparison. Overall, the performances of MT models were superior to those of ST models. The MT model using ANN algorithms showed the highest prediction accuracy with determination coefficients ranging from 0.704 to 0.886, root mean square errors between 0.223 and 0.410, and mean absolute errors ranging from 0.178 to 0.285 log units. Results showed that lipophilicity and polarizability of molecules significantly influence their partition behavior in organisms. Applicability domains (ADs) of the models were characterized by weighted molecular similarity density, and weighted inconsistency in molecular activities of structure-activity landscapes. When constrained by ADs, the models displayed enhanced predictive accuracy, making them valuable tools for the risk assessment and management of chemicals.

Multi-task machine learning of the electronic medical record to predict future symptoms among patients with cancer.

587 Background: Cancer and its treatments often cause symptoms. Automated warning systems could mitigate symptoms by alerting healthcare teams and enabling personalized preventative interventions. We developed a general-purpose longitudinal system for predicting symptomatic deterioration among outpatients undergoing intravenous systemic anti-cancer therapy. Methods: Patients treated for aerodigestive cancers at the Princess Margaret Cancer Centre were randomly divided into development and testing cohorts. For each treatment, machine learning was applied to preceding electronic medical record (EMR) data to predict patient-reported symptom deterioration, defined as at least a four point worsening on the Edmonton Symptom Assessment Scale. Features included diagnostic and treatment characteristics, laboratory tests, and patient-reported symptoms. Single-task (e.g., LASSO and XGboost) and multi-task (e.g., temporal CNNs, LSTM and Transformer) models were trained, tuned, and evaluated based on discrimination, calibration, and net benefit. Results: The cohort consisted of 3,998 patients who underwent 45,904 treatment sessions, with data across 400 features. Among these patients, 1,547 (38.6%) were female; median age was 64.0 (interquartile range 13.0). The most common diagnoses were lung (1,505, 37.6%), head and neck (696, 17.4%), and pancreatic cancers (685, 17.1%). The best model, a multi-task transformer, predicted symptom deterioration with an AUROC range of 0.732-0.822, marking a 1.4-6.2% improvement over the best single-task model. At a 10% alert rate, treatments associated with alerts would be enriched 4-13 fold for symptom deterioration (P&lt;0.001). The system was calibrated and would provide a net benefit across a wide range of threshold probabilities in decision curve analysis. Conclusions: Longitudinal general-purpose multi-task machine learning systems trained using EMR data can accurately predict a wide range of symptoms. Based on these results, automated warning systems for symptoms should be implemented and evaluated in real-time clinical practice to guide preventative interventions.[Table: see text]

Iterative multitask learning and inference from seismic images

SUMMARY Seismic interpretation aims to extract quantitative and interpretable attributes from a seismic image produced using some migration method to inform characteristics of a subsurface reservoir or target of interest. Current paradigms for computing seismic attributes mostly rely on single-task algorithms. We develop an iterative, multitask machine learning method to learn and infer multiple attributes from a seismic image. This method is composed of two stages: a multitask inference stage and a multimodal, multitask refinement stage. The basic mechanism of this method is that we train a multitask inference neural network to estimate a set of attributes, including a relative geological time volume, a denoised higher-resolution seismic image and multiple fault attributes (including probability, dip and strike), from a low-resolution, noisy seismic image; then we input the inferred attributes to a multitask refinement NN to enhance the raw inference results iteratively. The two multitask neural networks are trained separately based on synthetic seismic images and associated labels generated by geological modelling. Applications of this multitask learning and inference method to synthetic and field seismic images show that our method can improve the structural consistency among output seismic attributes compared with single-task neural networks, leading to more reliable automatic interpretation and subsurface characterization.

Monitoring canopy quality and improving equitable outcomes of urban tree planting using LiDAR and machine learning

Urban tree canopies are fundamental to mitigating the impacts of climate change within cities as well as providing a range of other important ecosystem, health, and amenity benefits. However, urban tree planting initiatives do not typically utilize data about both the horizontal and vertical dimensions of the tree canopy, despite height being a critical determinant of the quality and value of urban canopy cover. We present a novel pipeline that uses airborne LiDAR data to train a multi-task machine learning model to generate estimates of both canopy cover and height in urban areas. We apply this to multi-source multi-spectral imagery for the case study of Chicago, USA. Our results indicate that a multi-task UNet convolutional neural network can be used to generate reliable estimates of canopy cover and height from aerial and satellite imagery. We then use these canopy estimates to allocate 75,000 trees from Chicago’s recent green initiative under four scenarios, minimizing the urban heat island effect and then optimizing for an equitable canopy distribution, comparing results when only canopy cover is used, and when both canopy cover and height are considered. Through the introduction of this novel pipeline, we show that including canopy height within decision-making processes allows the distribution of new trees to be optimised to further reduce the urban heat island effect in localities where trees have the highest cooling potential and allows trees to be more equitably distributed to communities with lower quality canopies.

Detecting Social Contexts from Mobile Sensing Indicators in Virtual Interactions with Socially Anxious Individuals.

Mobile sensing is a ubiquitous and useful tool to make inferences about individuals' mental health based on physiology and behavior patterns. Along with sensing features directly associated with mental health, it can be valuable to detect different features of social contexts to learn about social interaction patterns over time and across different environments. This can provide insight into diverse communities' academic, work and social lives, and their social networks. We posit that passively detecting social contexts can be particularly useful for social anxiety research, as it may ultimately help identify changes in social anxiety status and patterns of social avoidance and withdrawal. To this end, we recruited a sample of highly socially anxious undergraduate students (N=46) to examine whether we could detect the presence of experimentally manipulated virtual social contexts via wristband sensors. Using a multitask machine learning pipeline, we leveraged passively sensed biobehavioral streams to detect contexts relevant to social anxiety, including (1) whether people were in a social situation, (2) size of the social group, (3) degree of social evaluation, and (4) phase of social situation (anticipating, actively experiencing, or had just participated in an experience). Results demonstrated the feasibility of detecting most virtual social contexts, with stronger predictive accuracy when detecting whether individuals were in a social situation or not and the phase of the situation, and weaker predictive accuracy when detecting the level of social evaluation. They also indicated that sensing streams are differentially important to prediction based on the context being predicted. Our findings also provide useful information regarding design elements relevant to passive context detection, including optimal sensing duration, the utility of different sensing modalities, and the need for personalization. We discuss implications of these findings for future work on context detection (e.g., just-in-time adaptive intervention development).