Yields Performance Improvements Research Articles

Abstract PO1-07-10: Blood Based Early Cancer Detection Assay

Abstract Background: Breast cancer screening programs utilizing mammography have been shown to be highly effective in identifying breast cancer in women over the age of 40. High breast density is an independent risk factor for breast cancer and makes mammograms more difficult to interpret, decreasing their sensitivity. In the spring of 2023, the FDA, who certifies all mammography facilities under the Mammography Quality Standards Act, updated its regulations to require that breast density status be reported to all individuals receiving a mammogram. The new guidelines now require individuals with dense breast tissue be notified of this status. The guidelines also recommend these women discuss additional screening options with their healthcare providers1. These additional screening options may include breast tomosynthesis, breast MRI, breast ultrasound, and/or molecular breast imaging. Many of these options require additional exposure to radiation, are expensive, and are not equitably available across the country. In addition, they all require a follow-up appointment. Compliance can be challenging given the top barriers to mammography cited include the need for transportation, child-care, and the ability to take time off from work2. Genece Health is developing a simpler and less expensive screening solution that identifies the presence of early to late-stage breast cancer using cfDNA from a single blood draw. The Genece Health assay utilizes an algorithm that leverages Artificial Intelligence and Machine Learning to analyze fragment size and end motif patterns in cfDNA as well as regional mutational density to detect presence of ctDNA originating from breast cancer. This algorithm provides highly sensitive and specific results in a preliminary data set. Methods: The preliminary data set, presented herein, is a cohort of over 50 retrospective breast cancer plasma samples and over 100 presumed normal samples. The breast cancer samples were collected at all stages of progression, from stage 0, or ductal carcinoma in situ (DCIS), through stage IV. The majority ( >50%) were from stage I breast cancer. 400 µL of double spun plasma, collected in Streck BCT devices, was processed to purify and isolate cfDNA. cfDNA was used to create WGS libraries that were sequenced on a NovaSeq 6000. Sequence data were analyzed using a bioinformatics pipeline that yields an ensemble probability that correlates to the presence or absence of ctDNA from breast cancer. Results: The Genece Health assay and algorithm performed with a specificity greater than 85%. With this specificity, the assay had a sensitivity greater than 85% in samples from stages II to IV and a slightly lower sensitivity in stage 0 and I samples. Follow-up analyses were conducted to stratify performance based on breast cancer type (e.g. invasive ductal carcinoma vs invasive lobular carcinoma) and HR, PR, and HER2 status (e.g. HER2-negatives vs HER2-positives). Conclusions: The presented preliminary data indicate that the Genece Health technology can be leveraged as a complement to mammography in indications, such as dense breast tissue, where there is an unmet need for an easy and cost-effective way to monitor for breast cancer. The ability to have a blood-based test to complement mammography could reduce the access barriers most cited by females in the United States. Follow-up studies with greater numbers of samples and additional training and optimizations of the algorithm will yield performance improvements that allow the assay to detect all types and stages of breast cancer.

Enhancing quantum annealing in digital–analog quantum computing

Digital–analog quantum computing (DAQC) offers a promising approach to addressing the challenges of building a practical quantum computer. By efficiently allocating resources between digital and analog quantum circuits, DAQC paves the way for achieving optimal performance. We propose an algorithm designed to enhance the performance of quantum annealing. This method employs a quantum gate to estimate the goodness of the final annealing state and find the ground state of combinatorial optimization problems. We explore two strategies for integrating the quantum annealing circuit into the DAQC framework: (1) state preparation, and (2) embedding within the quantum gate. While the former strategy does not yield performance improvements, we discover that the latter enhances performance within a specific range of annealing time. Algorithms demonstrating enhanced performance utilize the imaginary part of the inner product of two states from different quantum annealing settings. This measure reflects not only the energy of the classical cost function but also the trajectory of the quantum dynamics. This study provides an example of how processing quantum data using a quantum circuit can outperform classical data processing, which discards quantum information.

A dense and U-shaped transformer with dual-domain multi-loss function for sparse-view CT reconstruction.

CT image reconstruction from sparse-view projections is an important imaging configuration for low-dose CT, as it can reduce radiation dose. However, the CT images reconstructed from sparse-view projections by traditional analytic algorithms suffer from severe sparse artifacts. Therefore, it is of great value to develop advanced methods to suppress these artifacts. In this work, we aim to use a deep learning (DL)-based method to suppress sparse artifacts. Inspired by the good performance of DenseNet and Transformer architecture in computer vision tasks, we propose a Dense U-shaped Transformer (D-U-Transformer) to suppress sparse artifacts. This architecture exploits the advantages of densely connected convolutions in capturing local context and Transformer in modelling long-range dependencies, and applies channel attention to fusion features. Moreover, we design a dual-domain multi-loss function with learned weights for the optimization of the model to further improve image quality. Experimental results of our proposed D-U-Transformer yield performance improvements on the well-known Mayo Clinic LDCT dataset over several representative DL-based models in terms of artifact suppression and image feature preservation. Extensive internal ablation experiments demonstrate the effectiveness of the components in the proposed model for sparse-view computed tomography (SVCT) reconstruction. The proposed method can effectively suppress sparse artifacts and achieve high-precision SVCT reconstruction, thus promoting clinical CT scanning towards low-dose radiation and high-quality imaging. The findings of this work can be applied to denoising and artifact removal tasks in CT and other medical images.

Semantic-Aware Data Augmentation for Text-to-Image Synthesis

Data augmentation has been recently leveraged as an effective regularizer in various vision-language deep neural networks. However, in text-to-image synthesis (T2Isyn), current augmentation wisdom still suffers from the semantic mismatch between augmented paired data. Even worse, semantic collapse may occur when generated images are less semantically constrained. In this paper, we develop a novel Semantic-aware Data Augmentation (SADA) framework dedicated to T2Isyn. In particular, we propose to augment texts in the semantic space via an Implicit Textual Semantic Preserving Augmentation, in conjunction with a specifically designed Image Semantic Regularization Loss as Generated Image Semantic Conservation, to cope well with semantic mismatch and collapse. As one major contribution, we theoretically show that Implicit Textual Semantic Preserving Augmentation can certify better text-image consistency while Image Semantic Regularization Loss regularizing the semantics of generated images would avoid semantic collapse and enhance image quality. Extensive experiments validate that SADA enhances text-image consistency and improves image quality significantly in T2Isyn models across various backbones. Especially, incorporating SADA during the tuning process of Stable Diffusion models also yields performance improvements.

Logarithmic Learning Differential Convolutional Neural Network

Convolutional Neural Networks (CNNs) have revolutionized image classification through their innovative design and training methodologies in computer vision. Differential convolutional neural network with simultaneous multidimensional filter realization improved the performance of the convolutional neural network with calculation cost drawback. This paper introduces logarithmic learning integration into the differential Convolutional neural network to overcome the drawback by supplying faster error minimization and convergence. This task is done by incorporating LogRelu activation, a Logarithmic Cost Function, and unique logarithmic learning method. The effectiveness of the proposed approaches are evaluated by using various datasets and SGD/Adam optimizers. The first step is the adaptation of LogRelu activation function to convolutional and differential convolutional neural networks. The experiment results show that LogRelu integration to convolutional neural network and differential convolutional neural network yields performance improvements ranging from 1.61% to 5.44%. The same integration on ResNet-18, ResNet-34, and ResNet-50 enhances top-1 accuracy in the range of 3.07% and 9.96%. In addition to LogRelu activation function, a Logarithmic Cost Function with logarithmic learning method is also proposed and adapted to differential convolutional neural network. These improvements lead to a new differential convolutional neural network named as Logarithmic Differential Convolutional Neural Network (LDiffCNN), It consistently outperforms standard CNN by increasing the accuracy up to 3.02%. Notably, Logarithmic Differential Convolutional Neural Network demonstrates reduced training iterations up to 38% with faster convergence. The experimental results proved the efficiency of the proposed approach.

Category Alignment Mechanism for Few-Shot Image Classification.

While humans can excel at image classification tasks by comparing a few images, existing metric-based few-shot classification methods are still not well adapted to novel tasks. Performance declines rapidly when encountering new patterns, as feature embeddings cannot effectively encode discriminative properties. Moreover, existing matching methods inadequately utilize support set samples, focusing only on comparing query samples to category prototypes without exploiting contrastive relationships across categories for discriminative features. In this work, we propose a method where query samples select their most category-representative features for matching, making feature embeddings adaptable and category-related. We introduce a category alignment mechanism (CAM) to align query image features with different categories. CAM ensures features chosen for matching are distinct and strongly correlated to intra-and inter-contrastive relationships within categories, making extracted features highly related to their respective categories. CAM is parameter-free, requires no extra training to adapt to new tasks, and adjusts features for matching when task categories change. We also implement a cross-validation-based feature selection technique for support samples, generating more discriminative category prototypes. We implement two versions of inductive and transductive inference and conduct extensive experiments on six datasets to demonstrate the effectiveness of our algorithm. The results indicate that our method consistently yields performance improvements on benchmark tasks and surpasses the current state-of-the-art methods.

Self-supervised learning-based cervical cytology for the triage of HPV-positive women in resource-limited settings and low-data regime

Screening Papanicolaou test samples has proven to be highly effective in reducing cervical cancer-related mortality. However, the lack of trained cytopathologists hinders its widespread implementation in low-resource settings. Deep learning-assisted telecytology diagnosis emerges as an appealing alternative, but it requires the collection of large annotated training datasets, which is costly and time-consuming. In this paper, we demonstrate that the abundance of unlabeled images that can be extracted from Pap smear test whole slide images presents a fertile ground for self-supervised learning methods, yielding performance improvements compared to off-the-shelf pre-trained models for various downstream tasks. In particular, we propose Cervical Cell Copy-Pasting (C3P) as an effective augmentation method, which enables knowledge transfer from public and labeled single-cell datasets to unlabeled tiles. Not only does C3P outperforms naive transfer from single-cell images, but we also demonstrate its advantageous integration into multiple instance learning methods. Importantly, all our experiments are conducted on our introduced in-house dataset comprising liquid-based cytology Pap smear images obtained using low-cost technologies. This aligns with our long-term objective of deep learning-assisted telecytology for diagnosis in low-resource settings.

TSAR-ILP: Tile-Based, Synchronization-AwaRe ILP Allocating Heterogeneous Platforms for Streaming Applications

Automatic design space exploration (DSE) is key in hardware-software (HW/SW) co-design. To cope with the large design space, explorations are often heuristic-based and/or approximate yielding potentially locally optimal solutions. Without knowing the globally optimal solution, strong assertions about performance upper / lower bounds cannot be made. In contrast, integer linear programming (ILP) formulations can produce exact (optimal) solutions. Previous ILP-based formulations, however, lack support for tile-based architectures and realistic synchronization models, limiting their DSE capabilities. This work introduces a tile-based, synchronization-aware ILP (TSAR-ILP) formulation that overcomes previous limitations. With TSAR-ILP, the allocation / binding problems are introduced and formalized, attaining optimal solutions for mapping streaming applications onto template platforms. Using TSAR-ILP, this work explores a hardware acceleratorrich (HWACC-rich) platform with direct HWACC-to-HWACC communication under HW area constraints for 40 OpenVX applications. To illustrate design opportunities given by (a) the ILP formulation and (b) direct HWACC-to-HWACC communication, this paper analyzes the impact of job size. Results show that selecting smaller job sizes yields performance improvements and less area usage at the cost of slightly increased synchronization overhead. A job size reduction from 1 kB to 256 bytes gives 3.51x average performance increase across 40 applications. Finally, DSE with TSAR-ILP is shown not to be prohibitive through scalability analysis using a set of 5000 synthetic applications with varying size (10-125 nodes), with 94.3% of applications successfully achieving optimal solutions under 60 seconds.

Nighttime pedestrian detection based on Fore-Background contrast learning

The significance of background information is frequently overlooked in contemporary research concerning channel attention mechanisms. This study addresses the issue of suboptimal single-spectral nighttime pedestrian detection performance under low-light conditions by incorporating background information into the channel attention mechanism. Despite numerous studies focusing on the development of efficient channel attention mechanisms, the relevance of background information has been largely disregarded. By adopting a contrast learning approach, we reexamine channel attention with regard to pedestrian objects and background information for nighttime pedestrian detection, resulting in the proposed Fore-Background Contrast Attention (FBCA). FBCA possesses two primary attributes: (1) channel descriptors form remote dependencies with global spatial feature information; (2) the integration of background information enhances the distinction between channels concentrating on low-light pedestrian features and those focusing on background information. Consequently, the acquired channel descriptors exhibit a higher semantic level and spatial accuracy. Experimental outcomes demonstrate that FBCA significantly outperforms existing methods in single-spectral nighttime pedestrian detection, achieving state-of-the-art results on the NightOwls and TJU-DHD-pedestrian datasets. Furthermore, this methodology also yields performance improvements for the multispectral LLVIP dataset. These findings indicate that integrating background information into the channel attention mechanism effectively mitigates detector performance degradation caused by illumination factors in nighttime scenarios.

Artificial Bee Colony Based Iterative Heuristic Signal Detection Technique for Optical Wireless Communications

In this paper, we consider a problem of optical signal detection and present a novel artificial bee colony (ABC) based heuristic approach for optical spectrum sensing in a free-space optical wireless network that overcomes channel impairment due to noise uncertainty. In particular, we propose a generic ABC-based optimization framework in conjunction with eigenvalue-based detection (EVD). The uniqueness of the proposed algorithm is that it is executed iteratively by comparing various local solutions until an optimum solution is obtained. The channel is characterized as correlated multivariate fading and the received signal is quantified as a continuous waveform wideband detector. We derive the joint statistic of the correlated multivariate channel fading in the multiple-input multiple-output (MIMO) optical network. Subsequently, the marginal probability density function of the channel coefficients is obtained. For comparison purposes with existing optical spectrum sensing techniques, noise uncertainty is considered. It is demonstrated from simulations that the proposed ABC-EVD spectrum sensing yields performance improvement and the analytical results derived are verified. From the comparative study, it is also found that the proposed scheme outperforms existing optical spectrum sensing techniques, particularly in a regime of low signal-to-noise ratio. When compared to EVD, it is shown that, at a received signal-to-noise ratio of -15 dB, the performance gain achieved in terms of the probability of detection is 15% higher over the entire range of target probability of false alarm. When compared with the ED technique, the proposed technique outperforms it with approximately 50% higher probability of detection at the noise uncertainty of 1 dB and received SNR of -10 dB. The proposed algorithm offers fast convergence, strong robustness, and high flexibility.

Multivariate analysis of PET pharmacokinetic parameters improves inferential efficiency

PurposeIn positron emission tomography quantification, multiple pharmacokinetic parameters are typically estimated from each time activity curve. Conventionally all but the parameter of interest are discarded before performing subsequent statistical analysis. However, we assert that these discarded parameters also contain relevant information which can be exploited to improve the precision and power of statistical analyses on the parameter of interest. Properly taking this into account can thereby draw more informative conclusions without collecting more data.MethodsBy applying a hierarchical multifactor multivariate Bayesian approach, all estimated parameters from all regions can be analysed at once. We refer to this method as Parameters undergoing Multivariate Bayesian Analysis (PuMBA). We simulated patient–control studies with different radioligands, varying sample sizes and measurement error to explore its performance, comparing the precision, statistical power, false positive rate and bias of estimated group differences relative to univariate analysis methods.ResultsWe show that PuMBA improves the statistical power for all examined applications relative to univariate methods without increasing the false positive rate. PuMBA improves the precision of effect size estimation, and reduces the variation of these estimates between simulated samples. Furthermore, we show that PuMBA yields performance improvements even in the presence of substantial measurement error. Remarkably, owing to its ability to leverage information shared between pharmacokinetic parameters, PuMBA even shows greater power than conventional univariate analysis of the true binding values from which the parameters were simulated. Across all applications, PuMBA exhibited a small degree of bias in the estimated outcomes; however, this was small relative to the variation in estimated outcomes between simulated datasets.ConclusionPuMBA improves the precision and power of statistical analysis of PET data without requiring the collection of additional measurements. This makes it possible to study new research questions in both new and previously collected data. PuMBA therefore holds great promise for the field of PET imaging.

GF14f gene is negatively associated with yield and grain chalkiness under rice ratooning.

Ratoon rice cropping has been shown to provide new insights into overcoming the current challenges of rice production in southern China. However, the potential mechanisms impacting yield and grain quality under rice ratooning remain unclear. In this study, changes in yield performance and distinct improvements in grain chalkiness in ratoon rice were thoroughly investigated, using physiological, molecular and transcriptomic analysis. Rice ratooning induced an extensive carbon reserve remobilization in combination with an impact on grain filling, starch biosynthesis, and ultimately, an optimization in starch composition and structure in the endosperm. Furthermore, these variations were shown to be associated with a protein-coding gene: GF14f (encoding GF14f isoform of 14-3-3 proteins) and such gene negatively impacts oxidative and environmental resistance in ratoon rice. Our findings suggested that this genetic regulation by GF14f gene was the main cause leading to changes in rice yield and grain chalkiness improvement of ratoon rice, irrespective of seasonal or environmental effects. A further significance was to see how yield performance and grain quality of ratoon rice were able to be achieved at higher levels via suppression of GF14f.

Approximate execution and grouping of critical sections for performance‐accuracy tradeoff

SummaryApproximate computing enhances the performance and energy efficiency of applications, while still achieving acceptable accuracy. Some of the multithreaded applications can tolerate the accuracy loss when critical sections are approximately executed, which in turn will eliminate the synchronization overhead of these applications and increase their performance. In this study, our objective is to explore the behavior of the critical sections and selectively skip the ones yielding performance improvements with an acceptable accuracy loss. We have observed the behavior of 62 critical sections of 4 selected applications. We have grouped them depending on their effects on program execution and skipped or approximated them accordingly. Our experimental study indicates that skipping 76% of the critical sections offers 2.5 performance gain with 16% accuracy loss for Raytrace whereas 1.4 performance improvement with 17% accuracy loss is obtained for Radiosity on average when 36% of the critical sections are skipped. For Water_NSquared the performance gain is 1.1 with 9% accuracy loss on average with 79% of critical sections skipped. For the Ocean_CP application we see a performance improvement of 1.6 with accuracy loss 1% when we skip 38% of critical sections.

Distributed fixed-time sliding mode control of time-delayed quadrotors aircraft for cooperative aerial payload transportation: Theory and practice

This paper investigates a cooperative fixed-time control for networked-delayed and disturbed quadrotors in payload transportation missions. First, the input time-delayed tracking error system is adequately transformed into a delay-free system via Artstein’s reducing transformation. Second, a novel Distributed Cooperative Fixed-Time Stable Control Protocol (DCFTSCP) is proposed for the multi-quadrotor system to stabilize the geometrical spatial formation while carrying the payload. Furthermore, a disturbance observer is incorporated into the control framework to counteract the lumped disturbances and achieve the transportation task with a minimal swing. Overall, the designed DCFTSCP is able to stabilize the delayed tracking states at the origin in fixed-time uniformly to the Initial Conditions (ICs). It is worth mentioning that the proposed controller belongs to the homogeneous sliding mode control class. The reported related works rely on the bi-limit approximation method to prove the fixed-time convergence of the states. Nevertheless, this method suffers from a crucial shortcoming where it is unable mathematically to provide an explicit expression on the settling (convergence) time. In contrast, in the present study, the stability analysis is thoroughly investigated based on the algebraic Lyapunov tools, namely, Algebraic Lyapunov Equation (ALE) and Lyapunov Quadratic Function (LQF). Consequently, the upper-bound on the convergence-time is explicitly derived for the first time for time-delayed quadrotors aircraft used in cooperative payload transportation. MATLAB® simulations and real flight experiments are conducted to corroborate the theoretical studies of the paper. Overall, the results show that the proposed control protocol yields performance improvement regarding fixed-time tracking stability featuring fast transient, strong robustness, and high steady-state precision. In addition, the undesirable chattering problem of regular linear sliding mode control is noticeably attenuated. Furthermore, unlike conventional terminal sliding mode control techniques, the control input is singularity-free. Finally, several challenging issues in cooperative fixed-time control for networked-delayed quadrotors are highlighted for future research.

1-D Modeling of an Electroactive Electrolyte for Advanced Primary and Secondary Li Batteries

In the search for more energy- and power- dense battery materials, the cathode is most often the component that is engineered to boost operating voltage and capacity. However, new approaches to battery design have yielded performance improvements by designing electrolytes with reactive species that increase capacity and energy density of full-cells. These approaches borrow insights from commercial battery technologies such as Pb-H2SO4, Li-SO2, or Li-SOCl2, where the cathodic reaction involves liquid-phase active species1–3, while employing solid-phase cathode materials.To better understand how specific examples of electroactive electrolytes improve rather than impede performance, our work focuses on two battery chemistries: a primary Li-CFx cell with an irreversible electrolyte reaction, and a secondary Li-LMR cell with a reversible electrolyte reaction. For each case, we develop a 1-D COMSOL model to investigate how the reaction of electrolyte species affects cell impedance, impacts Li cation transport, alters the electrolyte conductivity, and finally how each of these mechanistically impact the energy and power density of the cell. Our modeling work is complemented by experimental studies that measure important electrolyte and cathode properties at various states of discharge or cycling.(1) Park, H.; Lim, H.-D.; Lim, H.-K.; Seong, W. M.; Moon, S.; Ko, Y.; Lee, B.; Bae, Y.; Kim, H.; Kang, K. High-Efficiency and High-Power Rechargeable Lithium–Sulfur Dioxide Batteries Exploiting Conventional Carbonate-Based Electrolytes. Nat. Commun. 2017, 8 (1), 14989. https://doi.org/10.1038/ncomms14989.(2) Schlaikjer, C. R.; Goebel, F.; Marincic, N. Discharge Reaction Mechanisms in Li / SOCl2 Cells. J. Electrochem. Soc. 1979, 126 (4), 513–522. https://doi.org/10.1149/1.2129078.(3) Ruetschi, P. Aging Mechanisms and Service Life of Lead–Acid Batteries. J. Power Sources 2004, 127 (1–2), 33–44. https://doi.org/10.1016/j.jpowsour.2003.09.052.

A Comparison of the Effects of Short-Term Physical and Combined Multi-Modal Training on Cognitive Functions.

Physical training has beneficial effects not only on physical fitness, but also on cognitive functions. The most effective way to improve cognitive functions via physical training as well as the degree to which training effects transfer to untrained cognitive functions is still unclear, however. Here, we investigated the effects of adaptive and multi-modal short-term training interventions on cognitive training gains and transfer effects. Over a period of 12 weeks, 102 employees of a car manufacturing company (age range 20 to 61 years) received trainer-guided exercises, consisting of either two adaptive training interventions, physical (strength) training and multi-modal (motor–cognitive) training, or non-adaptive strength training (active control group). For the multi-modal intervention, the “Agility Board” was employed, a novel, multi-modal training device. Pre- and post-training, psychometric tests were conducted to measure cognitive abilities, such as perceptual speed, attention, short-term memory, working memory, inhibition, and mental rotation. In addition, motor–cognitive performance was assessed. Compared with the active control group, both training groups showed enhanced performance at posttest. While multi-modal training yielded performance improvements only in trained tasks, physical training was associated with improvements in untrained working memory updating and immediate recall tasks, suggesting transfer effects to short-term and working memory functioning. In summary, the results demonstrate the importance of adaptive difficulty settings for short-term physical training interventions, at least for the enhancement of working memory.

Turning visions into results: unraveling the distinctive paths of leading with vision and autonomy to goal achievement

PurposeLeading with vision while granting employees autonomy is one effective organizational response to the demands of a dynamic external environment. The former is thought to align followers' behavior by providing guidance, the latter to increase variance in their behavior by relinquishing control; both exert beneficial but distinct effects on organizational performance. What has remained uncharted heretofore is how these leader behaviors shape their followers' cognition and, subsequently, yield improvements in performance. The authors argue that a leader's vision communication transforms followers' cognitive representation of their work. This not only enables them to specify their goals in alignment with the vision (goal clarity) but also to locate the meaning of their work within the bigger picture of the vision (construal level). By contrast, perceived autonomy in terms of power-sharing might directly affect followers' work engagement more narrowly.Design/methodology/approachThe authors tested the model on a sample of 408 employees from eleven enterprises of a holding company. In the survey, employees reported perceived vision communication and autonomy provided by their leader. Furthermore, the authors assessed the employees' goal attainment. To capture how employees represent their daily work activities, the authors measured their construal level and their goal clarity.FindingsThe results show that both perceived vision communication and granted autonomy improve employees' goal achievement. Moreover, two processes mediate the relationship between vision communication and goal achievement in followers: first, specifying goals in terms of clarity; second, composing a higher-level mental construal of their work. In contrast, no mediation of empowering leader behaviors was found.Originality/valueBetter goal achievement through visionary leadership is therefore achieved through cognitive alignment of followers, while leader-granted autonomy acts as a motivational tool directly on performance.

Simplified, interpretable graph convolutional neural networks for small molecule activity prediction

We here present a streamlined, explainable graph convolutional neural network (gCNN) architecture for small molecule activity prediction. We first conduct a hyperparameter optimization across nearly 800 protein targets that produces a simplified gCNN QSAR architecture, and we observe that such a model can yield performance improvements over both standard gCNN and RF methods on difficult-to-classify test sets. Additionally, we discuss how reductions in convolutional layer dimensions potentially speak to the “anatomical” needs of gCNNs with respect to radial coarse graining of molecular substructure. We augment this simplified architecture with saliency map technology that highlights molecular substructures relevant to activity, and we perform saliency analysis on nearly 100 data-rich protein targets. We show that resultant substructural clusters are useful visualization tools for understanding substructure-activity relationships. We go on to highlight connections between our models’ saliency predictions and observations made in the medicinal chemistry literature, focusing on four case studies of past lead finding and lead optimization campaigns.

Observer-based fixed-time continuous nonsingular terminal sliding mode control of quadrotor aircraft under uncertainties and disturbances for robust trajectory tracking: Theory and experiment

This paper solves an accurate fixed-time attitude and position control problems of a quadrotor UAV system. The aircraft system is subject to nonlinearities, parameter uncertainties, unmodeled dynamics, and external time-varying disturbances. To deal with the under-actuation problem of the quadrotor’s dynamics, a hierarchical control structure with an inner–outer loop framework is adopted for the flight control system design. Robust nonlinear control strategies for attitude and position control are innovatively proposed based on a new continuous nonsingular terminal sliding mode control (CNTSMC) scheme. A full-order homogeneous terminal sliding surface is designed for the attitude and position states in such a way that the sliding motion is fixed-time stable independently of the system’s initial condition. Hence, this contributes to enhancing the control system robustness. A disturbance observer-based control (DOBC) approach is developed to stabilize the inner rotational subsystem (attitude-loop). This compounded control structure integrates a finite-time observer (FTO) and the CNTSMC scheme. The FTO observer is incorporated into the control framework to cope with the strong perturbations. An output-feedback control approach is adopted for the outer translational subsystem (position-loop) to ensure a velocity-free control. In this context, the CNTSMC scheme is combined with a fixed-time extended state observer (FXESO) to achieve an active disturbance rejection control (ADRC) by estimating and canceling the lumped disturbances. Therefore, within the developed control approach including the robust CNTSMC scheme, DOBC, and ADRC strategies, robust and accurate trajectory tracking control can be achieved despite uncertainties and disturbances. Stability analysis of the closed-loop system is rigorously investigated by using the Lyapunov theorem, bi-limit homogeneous theory, and the notion of input-to-state stability (ISS). Extensive experimental tests under the influence of various disturbances are conducted to corroborate the theoretical findings. To this end, an effective model-based design (MBD) framework is established to implement the developed control algorithms in real autopilot hardware. Furthermore, processor-in-the-loop (PIL) experiments are also carried out within the MBD framework. A comparative study is made involving our control algorithms and other control strategies. Overall, the obtained results show that the synthesized control system yields performance improvement regarding fixed-time tracking stability featuring fast transient, strong robustness, and high steady-state precision. Besides, the chattering effect of regular linear sliding mode control (LSMC) is significantly alleviated. Moreover, unlike conventional TSMC, the control input shows no singularity.

A holding control strategy for diverging bus lines

Holding has been extensively used as control strategy to regulate public transport operations, especially to maintain even headways and prevent buses of the same line to bunch up. Applying holding to multiple lines requires however to deal with the transition between corridor and branching segments. In this study, we introduce a holding criterion for network configurations with lines that operate jointly along a common corridor and then diverge to individual branches serving different urban areas. The proposed holding decision rule accounts for all different passenger groups in the overlapping segment and considers the transition to individual line operation. The holding rule is evaluated using simulation for different demand levels and compositions and is compared with state-of-the-art control schemes for a real-world network. Results show that the proposed multi-line control yields performance improvements along the shared transit corridor as well as at the line level. The performance of the control scheme is affected by the demand composition and we provide indications regarding the conditions under which multi-line control is advisable.