Training Pipeline Research Articles

DLP: towards active defense against backdoor attacks with decoupled learning process

Deep learning models are well known to be susceptible to backdoor attack, where the attacker only needs to provide a tampered dataset on which the triggers are injected. Models trained on the dataset will passively implant the backdoor, and triggers on the input can mislead the models during testing. Our study shows that the model shows different learning behaviors in clean and poisoned subsets during training. Based on this observation, we propose a general training pipeline to defend against backdoor attacks actively. Benign models can be trained from the unreliable dataset by decoupling the learning process into three stages, i.e., supervised learning, active unlearning, and active semi-supervised fine-tuning. The effectiveness of our approach has been shown in numerous experiments across various backdoor attacks and datasets.

The leaky pipeline of diverse race and ethnicity representation in academic science and technology training in the United States, 2003-2019.

Diverse race and ethnicity representation remains lacking in science and technology (S&T) careers in the United States (US). Due to systematic barriers across S&T training stages, there may be sequential loss of diverse representation leading to low representation, often conceptualized as a leaky pipeline. We aimed to quantify the contemporary leaky pipeline of S&T training in the US. We analyzed US S&T degree data, stratified by sex and then by race or ethnicity, obtained from survey data the National Science Foundation and the National Center for Science and Engineering Statistics. We assessed changes in race and ethnicity representation in 2019 at two major S&T transition points: bachelor to doctorate degrees (2003-2019) and doctorate degrees to postdoctoral positions (2010-2019). We quantified representation changes at each point as the ratio of representation in the later stage to earlier stage (representation ratio [RR]). We assessed secular trends in the representation ratio through univariate linear regression. For 2019, the survey data included for bachelor degrees, 12,714,921 men and 10.612,879 women; for doctorate degrees 14,259 men and 12,860 women; and for postdoctoral data, 11,361 men and 8.672 women. In 2019, we observed that Black, Asian, and Hispanic women had comparable loss of representation among women in the bachelor to doctorate transition (RR 0.86, 95% confidence interval [CI] 0.81-0.92; RR 0.85, 95% CI 0.81-0.89; and RR 0.82, 95% CI 0.77-0.87, respectively), while among men, Black and Asian men had the greatest loss of representation (Black men RR 0.72, 95% CI 0.66-0.78; Asian men RR 0.73, 95% CI 0.70-0.77)]. We observed that Black men (RR 0.60, 95% CI 0.51-0.69) and Black women (RR 0.56, 95% CI 0.49-0.63) experienced the greatest loss of representation among men and women, respectively, in the doctorate to postdoctoral transition. Black women had a statistically significant decrease in their representation ratio in the doctorate to postdoctoral transition from 2010 to 2019 (p-trend = 0.02). We quantified diverse race and ethnicity representation in contemporary US S&T training and found that Black men and women experienced the most consistent loss in representation across the S&T training pipeline. Findings should spur efforts to mitigate the structural racism and systemic barriers underpinning such disparities.

Multilevel comparison of deep learning models for function quantification in cardiovascular magnetic resonance: On the redundancy of architectural variations.

Cardiac function quantification in cardiovascular magnetic resonance requires precise contouring of the heart chambers. This time-consuming task is increasingly being addressed by a plethora of ever more complex deep learning methods. However, only a small fraction of these have made their way from academia into clinical practice. In the quality assessment and control of medical artificial intelligence, the opaque reasoning and associated distinctive errors of neural networks meet an extraordinarily low tolerance for failure. The aim of this study is a multilevel analysis and comparison of the performance of three popular convolutional neural network (CNN) models for cardiac function quantification. U-Net, FCN, and MultiResUNet were trained for the segmentation of the left and right ventricles on short-axis cine images of 119 patients from clinical routine. The training pipeline and hyperparameters were kept constant to isolate the influence of network architecture. CNN performance was evaluated against expert segmentations for 29 test cases on contour level and in terms of quantitative clinical parameters. Multilevel analysis included breakdown of results by slice position, as well as visualization of segmentation deviations and linkage of volume differences to segmentation metrics via correlation plots for qualitative analysis. All models showed strong correlation to the expert with respect to quantitative clinical parameters (rz ' = 0.978, 0.977, 0.978 for U-Net, FCN, MultiResUNet respectively). The MultiResUNet significantly underestimated ventricular volumes and left ventricular myocardial mass. Segmentation difficulties and failures clustered in basal and apical slices for all CNNs, with the largest volume differences in the basal slices (mean absolute error per slice: 4.2 ± 4.5 ml for basal, 0.9 ± 1.3 ml for midventricular, 0.9 ± 0.9 ml for apical slices). Results for the right ventricle had higher variance and more outliers compared to the left ventricle. Intraclass correlation for clinical parameters was excellent (≥0.91) among the CNNs. Modifications to CNN architecture were not critical to the quality of error for our dataset. Despite good overall agreement with the expert, errors accumulated in basal and apical slices for all models.

State of research in adult Hospital Medicine: Updated results of a national survey and longitudinal analysis of national data.

We sought to understand the current state of research in adult Hospital Medicine by repeating a 2018 survey of leaders in Hospital Medicine with changes to improve the response rate of surveyed programs. We also analyzed the public sources of federal research funding and MEDLINE-indexed publications from 2010 through 2019 among members of the Society of Hospital Medicine (SHM). Of the 102 contacted leaders of Hospital Medicine groups across the country, 49 responded, for a total response rate of 48%. Among the 3397 faculty members represented in responding programs, 72 (2%) of faculty were identified as conducting research for more than 50% of their time. Respondents noted difficulties at every stage of the research development pipeline, from a lack of mentors to running a fellowship program to a lack of applicants seeking further research training. Improvements to our research training pipeline will be essential to the long-term improvement of our profession.

Vision-based automatic order check method for online medicine dispensing cabinet under incomplete data

In recent years, computer vision technology has been increasingly used in the retail industry, especially in unmanned retail. In the online medicine dispensing cabinet where the order images are captured along with the customer’s online order information, an automatic order check method is needed to verify whether the picked items match the order information and avoid the mis-dispensing of medicines. Well-built images dataset and annotations are needed for the task. However, it is difficult to construct datasets with fine annotations and keep the images up-to-date in the application scenario. Automatic order check is facing the problems of small samples, diverse and dynamic categories, weak supervision, and domain gaps. To deal with the order check task under incomplete data in real scenarios, we carried out the following works: (1) We proposed a framework for automatic order check (AOC), and collected an incomplete dataset in natural scenes. (2) We proposed a foreground detection method based on conventional image processing. Image matching methods based on local descriptors and text recognition were designed for classification. (3) We proposed a data simulation pipeline for rotated object detection model training with weak supervision. Our methods showed promising results on our challenging dataset.

168 Community Engagement, One Mile High: Developing a pipeline for training in community-based participatory research for investigators in Colorado

OBJECTIVES/GOALS: Describe the pipeline of training, coaching, and council opportunities through the Partnerships of Academicians and Communities for Translation (PACT) and how it enhances Community-Based Participatory Research (CBPR) practices and increases community participation and capacity. METHODS/STUDY POPULATION: We conceptualized a pipeline for building capacity for community-based participatory research (CBPR). Key components include Colorado Immersion Training in Community Engagement which introduces academic investigators to specific geographic and demographic communities in Colorado, Partnership Development grants that fund time to building relationships between research and community, Joint Pilot Grants to provide subsequent funding for collaborative research, and the PACT, consisting of Community Research Liaisons (CRLs) and Community and Academic Council Members. RESULTS/ANTICIPATED RESULTS: The pilot grants program was created in 2008; to date has awarded 138 partnerships. In response to academic and community needs, our core developed CIT in 2010. 16 CIT Participants received Partnership Development pilot grant funding, six of whom went on to receive Joint Pilot funding, and an additional 8 who started with a Joint Pilot award. There have been 24 Partnership Development awardees who received subsequent Joint Pilot funding. Ten CIT participants have either become a PACT council member, pilot grants reviewer, or PACT trainer or coach. There has been one person to complete the entire pipeline from CIT, pilot grant awardee, to PACT council member. DISCUSSION/SIGNIFICANCE: The pipeline is a strong foundation for engaging with academic and community researchers who aim to improve CBPR in Colorado. Our signature programs, CIT and Pilot Grants, are unique opportunities to increase community engagement across sectors and improve research practices.

MP25-20 TRENDS IN THE GLOBAL ENDOUROLOGY TRAINING PIPELINE: CONTEMPORARY ANALYSIS OF FELLOWSHIP MATCH OUTCOMES

You have accessJournal of UrologyCME1 Apr 2023MP25-20 TRENDS IN THE GLOBAL ENDOUROLOGY TRAINING PIPELINE: CONTEMPORARY ANALYSIS OF FELLOWSHIP MATCH OUTCOMES Garrett Ungerer, Reza Roshandel, Aaron Potretzke, and Kevin Koo Garrett UngererGarrett Ungerer More articles by this author , Reza RoshandelReza Roshandel More articles by this author , Aaron PotretzkeAaron Potretzke More articles by this author , and Kevin KooKevin Koo More articles by this author View All Author Informationhttps://doi.org/10.1097/JU.0000000000003253.20AboutPDF ToolsAdd to favoritesDownload CitationsTrack CitationsPermissionsReprints ShareFacebookLinked InTwitterEmail Abstract INTRODUCTION AND OBJECTIVE: The proportion of urology residents who choose to pursue fellowship training has steadily increased over the past decade. The Endourological Society fellowship match allocates fellowship positions for advanced endoscopic and minimally invasive surgical training in 83 programs across 5 continents. To understand contemporary trends in the global endourology training pipeline, we aimed to analyze results of the annual endourology fellowship match, conducted by the American Urologic Association. METHODS: We analyzed endourology fellowship match statistics from the AUA for the match years 2014-2023 to characterize trends in application volumes, vacancies, applicants, and successful matches. RESULTS: Over the past 10 years, the number of applicants entering the match increased from 54 to 86. In parallel with this increasing demand in fellowship training, the number if training opportunities has increased from 34 programs offering 36 positions in the 2014 match to 54 programs offering 59 positions in the 2023 match. Despite consistent excess of applicants relative to available positions, an average of 13 positions are unfilled each year (range 8-19) corresponding to 26% of available positions (range 18-43%). Of the potential reasons for this, applicant desirability did not appreciably change: an average of 75% of applicants (range 72%-80%) were ranked by programs each year. Increase in positions and applicants is occurring in parallel. For the prior 10 match cycles, programs received an average of 15 applications (range 14-18) and offered 7 interviews to applicants (range 6-9). Applicants consistently completed an average of 4 interviews (range 3-6). Finally, on average 75% of vacancies were filled by US graduating residents (range 55%-89%), compared to 17% by non-US-trained candidates and 8% by prior US graduates. The number of successful female candidates averaged 5 per year, increasing from 3 women in 2014 to 11 in 2023. CONCLUSIONS: Despite increasing applicant demand and global supply of endourology fellowship positions, only 3 out of 4 applicants are ranked by programs, and a relatively high proportion of vacancies go unfilled each year. Female and non-US-trained candidates remain underrepresented. Strategies to bolster the endourology training pipeline merit increased attention. Source of Funding: none to report © 2023 by American Urological Association Education and Research, Inc.FiguresReferencesRelatedDetails Volume 209Issue Supplement 4April 2023Page: e350 Advertisement Copyright & Permissions© 2023 by American Urological Association Education and Research, Inc.MetricsAuthor Information Garrett Ungerer More articles by this author Reza Roshandel More articles by this author Aaron Potretzke More articles by this author Kevin Koo More articles by this author Expand All Advertisement PDF downloadLoading ...

Heat stroke burden and validity of wearable-derived core temperature estimation during elite military training

ABSTRACT Objectives Exertional heat stroke (EHS) remains a significant health concern while training in hot environments, prompting the development of noninvasive wearable technology for monitoring estimated core temperature (ECT). The objective of this study was to review the effectiveness of an ECT device during elite military training by providing a case series of individuals who developed EHS. Methods This is a retrospective study of EHS cases during high-intensity training while wearing the ECT device, Zephyr BioharnessTM. Data was collected from January 2021 through September 2021 at the Air Force Special Warfare Training pipeline in San Antonio, TX. Rectal temperatures of EHS diagnoses, defined by central nervous system (CNS) dysfunction and rectal temperature approaching or >40°C (104°F), were compared to ECT reading via Zephyr BioharnessTM. Incidence rates and psychometric properties were calculated using R package. Results A total of 47,058 daily peak ECT measurements were collected among 1,364 trainees. A total of 499 trainees flagged as potential EHS by Zephyr BioharnessTM reading >39.7°C (103.5°F). The incidence of confirmed EHS was 0.8/1000 person-months. Of the 10 confirmed EHS cases (9 males, 1 female; age = 23.4 ± 2.7 yrs; BMI = 25 ± 2; body fat = 13 ± 5%), 8 trainees had a peak ECT reading below 39.7°C which resulted in a sensitivity of 20%, specificity of 98.9%, positive likelihood ratio (LR) of 18.93 (95% Confidence Interval [CI] 5.5–65.6), and a negative LR of 0.81 (95% CI 0.6–1.1). Conclusion ECT had substantial false positive and negative rates. Further studies are needed to validate this technology in other populations, and the algorithm used in this device needs to be refined to better capture the environmental and physical requirements in the special operations population.

Deep learning-enabled segmentation of ambiguous bioimages with deepflash2

Bioimages frequently exhibit low signal-to-noise ratios due to experimental conditions, specimen characteristics, and imaging trade-offs. Reliable segmentation of such ambiguous images is difficult and laborious. Here we introduce deepflash2, a deep learning-enabled segmentation tool for bioimage analysis. The tool addresses typical challenges that may arise during the training, evaluation, and application of deep learning models on ambiguous data. The tool’s training and evaluation pipeline uses multiple expert annotations and deep model ensembles to achieve accurate results. The application pipeline supports various use-cases for expert annotations and includes a quality assurance mechanism in the form of uncertainty measures. Benchmarked against other tools, deepflash2 offers both high predictive accuracy and efficient computational resource usage. The tool is built upon established deep learning libraries and enables sharing of trained model ensembles with the research community. deepflash2 aims to simplify the integration of deep learning into bioimage analysis projects while improving accuracy and reliability.

Future directions in neuropsychology: Training, education, clinical practice, and advocacy for Asians and Asian Americans

Objective: Given the need for increased equity, justice, and inclusion in neuropsychology, this paper aimed to present an initial perspective on key areas of understanding necessary to provide ethically and culturally responsive services and training to Asians and Asian Americans. Method: We first reviewed the terms Asian and Asian American and established the large multitude of individuals these terms encompass. Second, a brief review of the foundations for Asian American psychology is provided to set the stage for the unique considerations when evaluating individuals of Asian descent. Lastly, the necessity of using the social justice lens in education and training pipelines needed to propel the field forward is emphasized. Conclusions: Overall, this paper reviewed key information to provide a foundational level of understanding regarding the nuances of working with persons of Asian descent in the field of neuropsychology.

Dynamic ConvNets on Tiny Devices via Nested Sparsity

This work introduces a new training and compression pipeline to build nested sparse convolutional neural networks (ConvNets), a class of dynamic ConvNets suited for inference tasks deployed on resource-constrained devices at the edge of the Internet of Things. A nested sparse ConvNet consists of a single ConvNet architecture, containing <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$N$ </tex-math></inline-formula> sparse subnetworks with nested weights subsets, like a Matryoshka doll, and can trade accuracy for latency at runtime, using the model sparsity as a dynamic knob. To attain high accuracy at training time, we propose a gradient masking technique that optimally routes the learning signals across the nested weight subsets. To minimize the storage footprint and efficiently process the obtained models at inference time, we introduce a new sparse matrix compression format with dedicated compute kernels that fruitfully exploit the characteristic of the nested weights subsets. Tested on image classification and object detection tasks on an off-the-shelf ARM-M7 microcontroller unit (MCU), nested sparse ConvNets outperform variable-latency solutions naively built assembling single sparse models trained as stand-alone instances, achieving 1) comparable accuracy; 2) remarkable storage savings; and 3) high performance. Moreover, when compared to state-of-the-art dynamic strategies, such as dynamic pruning and layer width scaling, nested sparse ConvNets turn out to be Pareto optimal in the accuracy versus latency space.

Breaking barriers through the digital workforce: Providing IT training and employment pipelines for ex-offenders

With incarcerations on a steady increase, the United States is positioned to be the country with the highest incarceration rates in the world. A large majority of those incarcerated will be released back into the community, and represent a significant portion of the working-age population. Every year, incarcerated individuals are released with the expectation of reintegrating back into society, but are often met with opposition when looking for employment due to a lack of skills and the stigma of a criminal history. Lack of employment is one of the primary causes of recidivism. With rapid advancements and the pervasiveness of technology in our daily lives, offering Information Technology (IT) training to previously incarcerated individuals is a logical strategy to increase access to more stable opportunities and meet the growing demand for more IT professionals. Access to IT training would help increase skill sets, knowledge, and competencies of previously incarcerated workers in preparation for re-entering the workforce. This research applies Social Exchange Theory to examine approaches to managing technology to increase the domestic skilled IT workforce in more effective ways; while also increasing diversity and access to opportunities for those who were incarcerated.

A mentorship and incubation program using project-based learning to build a professional bioinformatics pipeline in Kenya.

The demand for well-trained bioinformaticians to support genomics research continues to rise. Unfortunately, undergraduate training in Kenya does not prepare students for specialization in bioinformatics. Graduates are often unaware of the career opportunities in bioinformatics, and those who are may lack mentors to help them choose a specialization. The Bioinformatics Mentorship and Incubation Program seeks to bridge the gap by laying the foundation for a bioinformatics training pipeline using project-based learning. The program selects six participants through an intensive open recruitment exercise for highly competitive students to join the program for four months. The six interns undergo intensive training within the first one and a half months before being assigned to mini-projects. We track the progress of the interns weekly through code review sessions and a final presentation at the end of the four months. We have trained five cohorts, most of whom have secured master's scholarships within and outside the country and job opportunities. We demonstrate the benefit of structured mentorship using project-based learning in filling the training gap after undergraduate programs to generate well-trained bioinformaticians who are competitive in graduate programs and bioinformatics jobs.

NeuRegenerate: A Framework for Visualizing Neurodegeneration.

Recent advances in high-resolution microscopy have allowed scientists to better understand the underlying brain connectivity. However, due to the limitation that biological specimens can only be imaged at a single timepoint, studying changes to neural projections over time is limited to observations gathered using population analysis. In this article, we introduce NeuRegenerate, a novel end-to-end framework for the prediction and visualization of changes in neural fiber morphology within a subject across specified age-timepoints. To predict projections, we present neuReGANerator, a deep-learning network based on cycle-consistent generative adversarial network (GAN) that translates features of neuronal structures across age-timepoints for large brain microscopy volumes. We improve the reconstruction quality of the predicted neuronal structures by implementing a density multiplier and a new loss function, called the hallucination loss. Moreover, to alleviate artifacts that occur due to tiling of large input volumes, we introduce a spatial-consistency module in the training pipeline of neuReGANerator. Finally, to visualize the change in projections, predicted using neuReGANerator, NeuRegenerate offers two modes: (i) neuroCompare to simultaneously visualize the difference in the structures of the neuronal projections, from two age domains (using structural view and bounded view), and (ii) neuroMorph, a vesselness-based morphing technique to interactively visualize the transformation of the structures from one age-timepoint to the other. Our framework is designed specifically for volumes acquired using wide-field microscopy. We demonstrate our framework by visualizing the structural changes within the cholinergic system of the mouse brain between a young and old specimen.

An Efficient Data Augmentation Method for Automatic Modulation Recognition from Low-Data Imbalanced-Class Regime

The application of deep neural networks to address automatic modulation recognition (AMR) challenges has gained increasing popularity. Despite the outstanding capability of deep learning in automatic feature extraction, predictions based on low-data regimes with imbalanced classes of modulation signals generally result in low accuracy due to an insufficient number of training examples, which hinders the wide adoption of deep learning methods in practical applications of AMR. The identification of the minority class of samples can be crucial, as they tend to be of higher value. However, in AMR tasks, there is a lack of attention and effective solutions to the problem of Imbalanced-class in a low-data regime. In this work, we present a practical automatic data augmentation method for radio signals, called SigAugment, which incorporates eight individual transformations and effectively improves the performance of AMR tasks without additional searches. It surpasses existing data augmentation methods and mainstream methods for solving low-data and imbalanced-class problems on multiple datasets. By simply embedding SigAugment into the training pipeline of an existing model, it can achieve state-of-the-art performance on benchmark datasets and dramatically improve the classification accuracy of minority classes in the low-data imbalanced-class regime. SigAugment can be trained for uniform use on different types of models and datasets and works right out of the box.

Double Momentum Backdoor Attack in Federated Learning

&#x0D; &#x0D; &#x0D; Federated learning is conceived as a privacy- preserving framework that trains deep neural networks from decentralized data. However, its decentralized nature exposes new attack surfaces. The privacy guarantees of federated learning prevent us from inspecting local data and training pipelines. These restrictions rule out many common defenses against poisoning attacks, such as data sanitization and traditional anomaly detection methods. The most devastating attacks are usually the ones that corrupt the model without altering the performance of the main task. Backdoor attacks are prominent examples of adversarial attacks that often go unnoticed in the absence of sophisticated defenses. This paper sheds light on backdoor attacks in federated learning, where we aim to manipulate the global model to misclassify the samples belonging to a particular task while also maintaining high accuracy on the main objective. Unlike existing works, we adopted a novel approach that directly manipulates the gradients’ momentums to introduce the backdoor. Specifically, the double momentum backdoor attack computes two momentums separately based on malicious and original inputs and uses them to update the model. Via experimental evaluation, we demonstrate that our attack scenario is capable of introducing the backdoor while successfully evading detection.&#x0D; &#x0D; &#x0D;

Towards global scale segmentation with OpenStreetMap and remote sensing

Land Use Land Cover (LULC) segmentation is a famous application of remote sensing in an urban environment. Up-to-date and complete data are of major importance in this field. Although with some success, pixel-based segmentation remains challenging because of class variability. Due to the increasing popularity of crowd-sourcing projects, like OpenStreetMap, the need for user-generated content has also increased, providing a new prospect for LULC segmentation. We propose a deep-learning approach to segment objects in high-resolution imagery by using semantic crowdsource information. Due to satellite imagery and crowdsource database complexity, deep learning frameworks perform a significant role. This integration reduces computation and labor costs. Our methods are based on a fully convolutional neural network (CNN) that has been adapted for multi-source data processing. We discuss the use of data augmentation techniques and improvements to the training pipeline. We applied semantic (U-Net) and instance segmentation (Mask R-CNN) methods and, Mask R–CNN showed a significantly higher segmentation accuracy from both qualitative and quantitative viewpoints. The conducted methods reach 91% and 96% overall accuracy in building segmentation and 90% in road segmentation, demonstrating OSM and remote sensing complementarity and potential for city sensing applications.

Robot Navigation in Crowded Environments: A Reinforcement Learning Approach

For a mobile robot, navigation in a densely crowded space can be a challenging and sometimes impossible task, especially with traditional techniques. In this paper, we present a framework to train neural controllers for differential drive mobile robots that must safely navigate a crowded environment while trying to reach a target location. To learn the robot’s policy, we train a convolutional neural network using two Reinforcement Learning algorithms, Deep Q-Networks (DQN) and Asynchronous Advantage Actor Critic (A3C) and develop a training pipeline that allows to scale the process to several compute nodes. We show that the asynchronous training procedure in A3C can be leveraged to quickly train neural controllers and test them on a real robot in a crowded environment.

KD-PatchMatch: A Self-Supervised Training Learning-Based PatchMatch

Traditional learning-based multi-view stereo (MVS) methods usually need to find the correct depth value from a large number of depth candidates, which leads to huge memory consumption and slow inference. To address these problems, we propose a probabilistic depth sampling in the learning-based PatchMatch framework, i.e., sampling a small number of depth candidates from a single-view probability distribution, which achieves the purpose of saving computational resources. Furthermore, to overcome the difficulty of obtaining ground-truth depth for outdoor large-scale scenes, we also propose a self-supervised training pipeline based on knowledge distillation, which involves self-supervised teacher training and student training based on knowledge distillation. Extensive experiments show that our approach outperforms other recent learning-based MVS methods on DTU, Tanks and Temples, and ETH3D datasets.

Gaussian Process Regression Models for Predicting Atomic Energies and Multipole Moments.

Developing a force field is a difficult task because its design is typically pulled in opposite directions by speed and accuracy. FFLUX breaks this trend by utilizing Gaussian process regression (GPR) to predict, at ab initio accuracy, atomic energies and multipole moments as obtained from the quantum theory of atoms in molecules (QTAIM). This work demonstrates that the in-house FFLUX training pipeline can generate successful GPR models for six representative molecules: peptide-capped glycine and alanine, glucose, paracetamol, aspirin, and ibuprofen. The molecules were sufficiently distorted to represent configurations from an AMBER-GAFF2 molecular dynamics run. All internal degrees of freedom were covered corresponding to 93 dimensions in the case of the largest molecule ibuprofen (33 atoms). Benefiting from active learning, the GPR models contain only about 2000 training points and return largely sub-kcal mol-1 prediction errors for the validation sets. A proof of concept has been reached for transferring the model produced through active learning on one atomic property to that of the remaining atomic properties. The prediction of electrostatic interaction can be assessed at the intermolecular level, and the vast majority of interactions have a root-mean-square error of less than 0.1 kJ mol-1 with a maximum value of ∼1 kJ mol-1 for a glycine and paracetamol dimer.