Using Radar for Edge-based Live Learning

Computational skewness is a significant challenge in multi-tenant data-parallel clusters that introduce dynamic heterogeneity of machine capacity in distributed data processing. Previous efforts to addressing skewness mostly focus on batch jobs based on the assumption that processing time is linearly dependent on the size of partitioned data. However, they are illsuited for iterative machine learning (ML) jobs, which (1) exhibit a non-linear relationship between the size of partitioned parameters and processing time within each iteration, and (2) show an explicit binding relationship between input data and parameters for parameter update. In this paper, we present FlexPara, a parameter partition approach that leverages the non-linear relationship and provisions adaptive tasks to match the distinct machine capacity so as to address the skewness in iterative ML jobs on data-parallel clusters. FlexPara first predicts task processing time based on a capacity model designed for iterative ML jobs without the linear assumption. It then partitions parameters to parallel tasks through proactive parameter reassignment. Such reassignment can significantly reduce network transmission cost incurred by input data movement due to the binding relationship. We implement FlexPara in Spark and evaluate it with various ML jobs. Experimental results show that compared to hash partition, FlexPara speeds up the execution by up to 54% and 43% in private and NSF Chameleon clusters, respectively.

Elraglusib is a selective GSK-3 inhibitor that has activity on a broad array of neoplasms as well as their immune microenvironment. However the determinants of tumor response have not been elucidated. Here, we use genomic panels of cancer driver genes to predict patient response to elraglusib therapy as measured by RECIST criteria. Patient response in the Phase I 1801 trial (NCT03678883) was matched to genomic data from patient tumors, with 80 cases used to train Machine Learning models and 26 cases reserved as a test set for model evaluation. When predicting response from mutation status alone our models reached a modest accuracy of 65% in the test set, which we attributed to the sparsity of genomic data. To remedy this, we engineered pathway-based features to provide combinatorial information while ensuring feature stability. To select the optimal combination of features, we designed an iterative design process that first randomized feature combinations and then created refined models from the highest performing pathway features. This resulted in several final models with greater than 88% accuracy. We retained several models of similar accuracy to more comprehensively identify potential biomarkers and combinatorial relationships. These models were then interpreted using SHapley Additive exPlanations (SHAP) where we were able to identify highly predictive features for both the entire pan-cancer cohort and specific histologies. We found the feature constructed from the Reactome pathway “Chromatin Modifying Enzymes” frequently occurred in high performing models with a high-ranking SHAP value. This feature included genes such as SMARCA4, HDAC2, and KDM5A, and was negatively associated with patient response. Another frequently observed high-ranking pathway feature with a negative response association was based on the “Innate Immune System” Reactome pathway. Many of the features highlighted in our models were negatively associated with elraglusib response and thus could be general markers of poor prognosis. To address this, we altered our pipeline to test only positively associated features which substantially reduced our feature pool to only 285 features but still produced models with accuracies of up to 81%. Analysis of key features used by these models identified several positive markers of elraglusib including mutations in the POLE gene, which previously has been linked to DNA-damage response deficiency and anti-tumor immune response. In sum, we forecast elraglusib response for a variety of tumor histologies and simultaneously reveal potential mechanisms of elraglusib sensitivity and biological action. In future work we will investigate the utility of using POLE and other identified biomarker candidates to predict the likelihood of patient response to elraglusib, as well as whether our machine learning models will be an effective tool to guide patient stratification. Citation Format: Joseph McDermott, Taylor Weiskittel, Daniel Billadeau, Benedito Carneiro, Hu Li, Andrew Mazar. Elraglusib response prediction and mechanistic discovery using iterative machine learning. [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 5355.

Scientific workflows are one of the well-established pillars of modern large-scale computational science. More recently, scientists have started to leverage machine learning (ML) capabilities in their workflows, leading to a new category of scientific workflows, denoted as scientific ML workflows. ML is not only about training and inference, modern ML workflows also involve complex data processing steps before the training can start, which are not often accounted for in most performance studies. In this work, we consider scientific ML workflows, from data pre-processing to training, inference, and model evaluation. We aim to explore (i) how scientific ML workflows differ from more traditional scientific workflows and; (ii) how we can characterize ML workflows both in terms of execution time and data movements when executing on a target cloud platform. We select three representative workflows, ranging from image classification to natural language processing and image segmentation, which have been executed using the academic cloud platform, Chameleon. We build four realistic deployment scenarios for each workflow, which stress data movements during workflow executions. Then, we compare the performance observed when utilizing these different configurations and study how different settings impact overall workflows performance and efficiency when running on cloud infrastructures. Finally, we summarize our findings and discuss performance impacts when augmenting scientific workflows with ML techniques and how traditional workflow management systems can improve their support for such workflows.

Ruin has a bad name. Despite the obvious complications, failure provides a rich opportunity—how better to understand a building’s physicality than to watch it collapse? This paper offers a novel method to exploit failure through physical simulation and iterative machine learning. Using technology traditionally relegated to special effects, we can now understand collapse on a granular level: since modern-day physics engines track object-object collisions, they enable a close reading of the spatial preferences that underpin ruin. In the case of bricks, that preference is relatively simple—to fall. By idealizing bricks as rigid bodies, one can understand the effects of gravitational force on each individual brick in a masonry structure. These structures are sometimes able to ‘settle,’ resulting in a stable equilibrium state; in many cases, it means that they will simply collapse. Analyzing ruin in this way is informative, to be sure, but it proves most useful when applied in series. The evolutionary solver described in this paper closely monitors the performance of constituent bricks and ensures that the most successful structures are emulated by later generations. The tool consists of two parts: a user interface for design and the solver itself. Once the architect produces a potential design, the solver performs an evolutionary optimization; after a few hundred iterations, the end result is a structurally sound version of the unstable original. It is hoped that this hybrid of top-down and bottom-up design strategies offers an architecture that is ultimately strengthened by its contingencies.

Enabled by a rich ecosystem of Machine Learning (ML) libraries, programming using learned models , i.e., Software-2.0 , has gained substantial adoption. However, we do not know what challenges developers encounter when they use ML libraries. With this knowledge gap, researchers miss opportunities to contribute to new research directions, tool builders do not invest resources where automation is most needed, library designers cannot make informed decisions when releasing ML library versions, and developers fail to use common practices when using ML libraries. We present the first large-scale quantitative and qualitative empirical study to shed light on how developers in Software-2.0 use ML libraries, and how this evolution affects their code. Particularly, using static analysis we perform a longitudinal study of 3,340 top-rated open-source projects with 46,110 contributors. To further understand the challenges of ML library evolution, we survey 109 developers who introduce and evolve ML libraries. Using this rich dataset we reveal several novel findings. Among others, we found an increasing trend of using ML libraries: The ratio of new Python projects that use ML libraries increased from 2% in 2013 to 50% in 2018. We identify several usage patterns including the following: (i) 36% of the projects use multiple ML libraries to implement various stages of the ML workflows, (ii) developers update ML libraries more often than the traditional libraries , (iii) strict upgrades are the most popular for ML libraries among other update kinds, (iv) ML library updates often result in cascading library updates, and (v) ML libraries are often downgraded (22.04% of cases). We also observed unique challenges when evolving and maintaining Software-2.0 such as (i) binary incompatibility of trained ML models and (ii) benchmarking ML models. Finally, we present actionable implications of our findings for researchers, tool builders, developers, educators, library vendors, and hardware vendors.

Introduction: Aircrew equipment assemblies are life critical equipment worn by the aircrew to ensure protection from various life-threatening environmental conditions that may occur in nominal or off-nominal flight conditions. It is important for the sizing system of body worn aircrew equipment to be divided into different size classes such that it is as representative of existing groups in the population as feasible. Machine learning techniques provide a powerful tool to develop efficient sizing systems that are representative of actual body types existing in the population. The objective of the study was to use machine learning clustering techniques to identify naturally existing body types to formulate an optimal sizing system for an aircrew helmet. Material and Methods: The standard sizing schedule of an aircrew helmet using head length and head breadth was studied. An iterative semi-supervised machine learning protocol called K-means clustering was used to identify naturally occurring clusters of head types within the population. The cluster boundaries were identified to develop the final sizing schedule. Analysis of variance (ANOVA) with post hoc analysis was carried out on the four head lengths and head breadths. Results: Four clusters of head type were identified using K-means clustering. This led to a sizing schedule which can be descriptively called “Short-Narrow,” “Short-Broad,” “Long-Narrow,” “Long-Broad.” ANOVA showed a statistically significant difference between the four clusters for both head length and head breadth. Conclusion: Consideration of several variables in sizing schedules often fails to meet desired fitment. Standard sizing methods use linear partitions on individual control parameters. In reality, the proportional variation between parameters is not linear. Machine learning tools help in identification of naturally occurring clusters within the database considering several variables at the same time. This study uses machine learning techniques to identify existing groups in a population and develop a final sizing schedule. Similar analysis for other aircrew equipment assemblies can be carried out to determine sizing schedules which assist in design and fitment.

Machine learning application in the oil and gas industry is rapidly becoming popular and in recent years has been applied in the optimization of production for various reservoirs. The objective of this paper is to evaluate the efficacy of advanced machine learning algorithms in reservoir production optimization. A 3-D geological model was constructed based on permeability calculated using a machine learning technique which involved different architectures of algorithms tested using a 5-fold cross-validation to decide the best machine learning algorithm. Sensitivity analysis and a subsequent history matching were conducted using a machine learning workflow. The aquifer properties, permeability heterogeneity in different directions and relative permeability were the control variables assessed. Field development scenarios were exploited with the objective to optimize cumulative oil recovery. The impact of using a normal depletion plan to a secondary recovery plan using waterflooding was investigated. Different injection well placement locations, well patterns as well as the possibility of converting existing oil producing wells to water injection wells were exploited. Considering the outcome of an economic analysis, the optimum development strategy was realized as an outcome for the optimization process. Prior to forecasting cumulative oil production using artificial neural network (ANN) for the optimization process on the generated surrogate model, a sensitivity analysis was performed where the well location, injection rates and bottomhole pressure of both the producer and injector wells were specified as control variables. The water cut as part of the optimization process was utilized as a secondary constraint. Forecasting was performed for a 15-year period. The history-matching results from the constructed geological model showed that the oil rate, water rate, bottom hole pressure, and average reservoir pressure were matched within a 10% deviation from the observed data. In this study, the ANN optimizer was found to provide the best results for the field cumulative oil production. Using a secondary recovery development plan was observed to significantly increase the cumulative oil production. A machine learning based proxy model was built for the prediction of cumulative oil production to reduce computational time. In this study, we propose an approach applied to reservoir production optimization utilizing a machine learning workflow. This was accomplished by utilizing a surrogate model which was calibrated with a number of training simulations and then optimized using advanced machine learning algorithms. A detailed economic analysis was also conducted showing the impact of a variety of field development strategies.

In this work, we present a new machine learning-based framework for the accurate estimation of the shot-noise limited photon-counting Poisson channel by developing a state-of-art iterative unsupervised learning algorithm for intelligent optical communications. By accurate estimation, we mean that it achieves very low estimation error even when the effect of the received data is unpredictable. We consider a realistic situation where modulated symbols are assumed to be hidden or unobserved latent variables, thereby making conventional estimation algorithms based on maximum likelihood approach unsuitable or inefficient. In particular, we consider a probabilistic model and assume that the received data is not labeled. With this unpredictable data considered, a novel iterative machine learning framework is developed based on an expectation and maximization algorithm. The proposed algorithm avoids the need to choose an appropriate step size as required in gradient method based algorithms. It is shown that it significantly outperforms the least square and the Viterbi detection technique.

Machine learning (ML) has become an increasingly popular tool for use in neurosurgical research. The number of publications and interest in the field have recently seen significant expansion in both quantity and complexity. However, this also places a commensurate burden on the general neurosurgical readership to appraise this literature and decide if these algorithms can be effectively translated into practice. To this end, the authors sought to review the burgeoning neurosurgical ML literature and to develop a checklist to help readers critically review and digest this work. The authors performed a literature search of recent ML papers in the PubMed database with the terms "neurosurgery" AND "machine learning," with additional modifiers "trauma," "cancer," "pediatric," and "spine" also used to ensure a diverse selection of relevant papers within the field. Papers were reviewed for their ML methodology, including the formulation of the clinical problem, data acquisition, data preprocessing, model development, model validation, model performance, and model deployment. The resulting checklist consists of 14 key questions for critically appraising ML models and development techniques; these are organized according to their timing along the standard ML workflow. In addition, the authors provide an overview of the ML development process, as well as a review of key terms, models, and concepts referenced in the literature. ML is poised to become an increasingly important part of neurosurgical research and clinical care. The authors hope that dissemination of education on ML techniques will help neurosurgeons to critically review new research better and more effectively integrate this technology into their practices.

To broaden the accessibility of cell and gene therapies, it is essential to develop and optimize nonviral, cell type-preferential gene carriers such as lipid nanoparticles (LNPs). While high-throughput screening (HTS) approaches have proven effective in accelerating LNP discovery, they are often costly, labor-intensive, and do not consistently yield actionable design rules that direct screening efforts toward the most relevant chemical and formulation parameters. In this study, we employed a machine learning (ML) workflow, utilizing well-curated plasmid DNA LNP transfection data sets across six cell types, to extract compositional and chemical insights from HTS studies. Our approach achieved prediction errors averaging between 5 and 10%, depending on the cell type. By applying SHapley Additive exPlanations to our ML models, we uncovered key composition-function relationships that govern cell type-preferential LNP transfection efficiency. Notably, we identified consistent LNP composition parameters that enhance in vitro transfection efficiency across diverse cell types, including a helper lipid molar percentage of charged lipids between 9 and 50% and the inclusion of cationic/zwitterionic helper lipids. Additionally, several parameters were found to modulate cell type-preferentiality, such as the total molar percentage of ionizable and helper lipids, N/P ratio, PEGylated lipid molar percentage of uncharged lipids, and hydrophobicity of the helper lipid. This study leverages HTS of compositionally diverse LNP libraries combined with ML analysis to elucidate the interactions between lipid components in LNP formulations, providing insights that contribute to the design of LNP compositions tailored for cell type-preferential transfection.

Organic radical batteries (ORBs) based on the TEMPO (2,2,6,6-tetramethylpiperidin-1-yl oxyl) radical have drawn significant attention, owing to their unique redox properties. A key factor influencing ORB’s redox properties, i.e., the kinetics of the electron transfer between the TEMPO–TEMPO+ pairs, is the communication between the underlying redox-active states as given by the electronic coupling. However, due to the complex structure, predicting accurate electronic couplings for these pairs is computationally expensive and challenging. In this study, we introduce a machine learning (ML) workflow to predict the electronic coupling for TEMPO–TEMPO+ pairs simply by their specific geometric orientations. For the ML models, a data set was generated through time-dependent density functional theory calculations coupled with the Generalized Mulliken Hush method to assess energies, (transition-)dipole moment, and couplings for specific TEMPO–TEMPO+ configurations obtained from classical molecular dynamics simulations that mimic a realistic electrolyte environment. Our results demonstrate that, among the three ML models─linear regression, kernel ridge regression (KRR), and random forest─the KRR model, with its kernel-based approach, most effectively handles the correlated orientation-based descriptors. Moreover, our SHapley Additive exPlanations (SHAP)-based feature importance analysis indicates that multiple orientation factors jointly influence electronic coupling, rather than any single distance or angle dominating, with each parameter’s impact strongly contingent on the values of the others which is in agreement with previous studies computational by the consortium.

Complex stochastic systems require the control of their stochastic distributions. This keynote paper will address both modelling and control of such systems and will consist of the following aspects: 1) Neural network based modelling of the stochastic distribution systems; 2) Control framework for the stochastic profile control of the systems; 3) Iterative learning of the space variables so as to achieve a batch-by-batch improvement of the closed loop performance. The above will include our originated research on complex stochastic systems in terms of probability density function (pdfs) control, where neural networks such as RBF will be used to approximate the output pdfs and the system dynamics. This will be followed by the iterative machine learning for the RBF basis functions on a batch-by-batch basis so as to improve the closed loop performance both in the time and in the space. Applications to particle size distribution control and 3D paper web distribution control will be discussed in the presentation.

Traditional workflows in chemical engineering, from design to characterization, have served the field well. Emerging technologies, such as machine learning, now offer opportunities for faster design iterations and better integration of the various development stages. This study demonstrates how machine learning (ML) can reduce this gap by directly integrating experimental data into predictive models. Focusing on a Additive Manufactured helical micro-distillation column design for separating liquid mixtures, several ML models were trained to predict the number of theoretical stages based on geometric parameters and operating conditions. 11 variants were designed, built, and tested to collect 197 experimental data sets to assess the feasibility of five predictive ML models from experimental data. Among the algorithms tested, the Gradient Boosting Regressor achieved the best performance with a coefficient of determination R 2 = 0 . 7934 . The work highlights the behavior of the model across different regimes, identifies key sources of error, and proposes a hybrid experimental–ML workflow for rapid screening of distillation designs. This approach accelerates process development and reduces the need for extensive experimentation, especially in time-consuming tasks such as distillation. • ML predicts KPIs for complex AM helical distillation columns using real data. • Multiple supervised ML models benchmarked for AM-induced geometric effects. • Multi-output modeling enhances stability and captures physical correlations. • Hybrid ML–experimental workflow identifies high-value regions for testing. • Framework accelerates design and screening of AM-enabled separation units.

Applications of machine learning in metal-organic frameworks

Alterations in microbiome composition have been linked to many human diseases, including colorectal cancer or precancerous liver diseases. Findings from microbiome-association studies are increasingly explored as promising avenues for clinical applications due to their diagnostic or therapeutic potential. However, microbiome data complexity and the lack of integrated computational tools make it difficult to arrive at robust associations and predictive disease models. Here, we present the SIAMCAT R package as a modular and user-friendly toolbox for machine learning workflows, statistical analysis, and confounder detection for case-control microbiome datasets. Previously, we conducted a machine learning meta-analysis of colorectal cancer metagenomics studies (Wirbel, Pyl et al., Nat Med 2019) using the functionalities of SIAMCAT, which is now available to the wider community via Bioconductor. We showcase how SIAMCAT is a versatile tool by applying it to a large set of metagenomic datasets that have been processed with a wide variety of taxonomic and functional profiling tools. We furthermore demonstrate how SIAMCAT can help to detect confounding factors in microbiome association studies. By making stringent machine learning workflows and analysis pipelines available through a user-friendly and flexible interface, SIAMCAT will facilitate microbiome data analyses, biomarker discovery, and thus the translation of microbiome research to clinical applications, while simultaneously improving statistical rigor and safeguarding against common machine learning pitfalls. Citation Format: Jakob Wirbel, Konrad Zych, Morgan Essex, Nicolai Karcher, Ece Kartal, Guillem Salazar, Peer Bork, Shinichi Sunagawa, Georg Zeller. The SIAMCAT R package enables statistical and machine learning analyses for case-control microbiome datasets [abstract]. In: Proceedings of the AACR Special Conference on the Microbiome, Viruses, and Cancer; 2020 Feb 21-24; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2020;80(8 Suppl):Abstract nr A40.