Robustness Tradeoffs Research Articles

The Capacity and Robustness Trade-Off: Revisiting the Channel Independent Strategy for Multivariate Time Series Forecasting

The Capacity and Robustness Trade-Off: Revisiting the Channel Independent Strategy for Multivariate Time Series Forecasting

Trustworthy V2G scheduling and energy trading: A blockchain-based framework

The rapid growth of electric vehicles (EVs) and the deployment of vehicle-to-grid (V2G) technology pose significant challenges for distributed power grids, particularly in fostering trust and ensuring effective coordination among stakeholders. Establishing a trustworthy V2G operation environment is crucial for enabling large-scale EV user participation and realizing V2G’s potential in real-world applications. In this paper, an integrated scheduling and trading framework is developed to conduct transparent and efficacious coordination in V2G operations. In blockchain implementation, a cyber-physical blockchain architecture is proposed to enhance transaction efficiency and scalability by leveraging smart charging points (SCPs) for rapid transaction validation through a fast-path practical byzantine fault tolerance (fast-path PBFT) consensus mechanism. From the energy dispatching perspective, a game-theoretical pricing strategy is employed and smart contracts are utilized for autonomous decision-making between EVs and operators, aiming to optimize the trading process and maximize economic benefits. Numerical evaluation of blockchain consensus shows the effect of the fast-path PBFT consensus in improving systems scalability with a balanced trade-off in robustness. A case study, utilizing real-world data from the Southern University of Science and Technology (SUSTech), demonstrates significant reductions in EV charging costs and the framework’s potential to support auxiliary grid services.

Distance distributions and runtime analysis of perceptual hashing algorithms

Perceptual image hashing refers to a class of algorithms that produce content-based image hashes. These systems use specialized perceptual hash algorithms like Phash, Microsoft’s PhotoDNA, or Facebook’s PDQ to generate a compact digest of an image file that can be roughly compared to a database of known illicit-content digests. Time taken by perceptual hashing algorithms to generate hash code has been computed. Then, we evaluated perceptual hashing algorithms on two million dataset of images. The produced nine variants of the original images were computed and then several distances were calculated. There have been several studies in the past, but in the existing literature size of the data is small and there are very few studies with hash code computation time and robustness tradeoff. This work shows that existing perceptual hashing algorithms are robust for most of the content-preserving operations and there is a tradeoff between computation time and robustness.

A transient radial cortical microtubule array primes cell division in Arabidopsis

Although the formation of new walls during plant cell division tends to follow maximal tensile stress direction, analyses of individual cells over time reveal a much more variable behavior. The origin of such variability as well as the exact role of interphasic microtubule behavior before cell division have remained mysterious so far. To approach this question, we took advantage of the Arabidopsis stem, where the tensile stress pattern is both highly anisotropic and stable. Although cortical microtubules (CMTs) generally align with maximal tensile stress, we detected a specific time window, ca. 3 h before cell division, where cells form a radial pattern of CMTs. This microtubule array organization preceded preprophase band (PPB) formation, a transient CMT array predicting the position of the future division plane. It was observed under different growth conditions and was not related to cell geometry or polar auxin transport. Interestingly, this cortical radial pattern correlated with the well-documented increase of cytoplasmic microtubule accumulation before cell division. This radial organization was prolonged in cells of the trm678 mutant, where CMTs are unable to form a PPB. Whereas division plane orientation in trm678 is noisier, we found that cell division symmetry was in contrast less variable between daughter cells. We propose that this "radial step" reflects a trade-off in robustness for two essential cell division attributes: symmetry and orientation. This involves a "reset" stage in G2, where an increased cytoplasmic microtubule accumulation transiently disrupts CMT alignment with tissue stress.

An Information-Theoretic Approach to Morphosyntactic Complexity in English, Dutch and German

ABSTRACT Though equi-complexity of languages has long been an assumption in the study of language, recent research has argued that languages differ with regard to their morphological complexity. Some languages rely more on bound morphology to express grammatical meanings, whereas other languages rely more on lexical or word-order-based strategies. It is a moot point to what extent these differences correlate with demographic factors such as population size and language contact, and how complexity should be measured. In this study, we use information-theory, more specifically Kolmogorov Complexity, to assess morphological and word-order-based strategies and apply the procedure to three West Germanic languages, English, Dutch and German, which have been argued to form a continuum along the morphological complexity cline, plausibly due to different rates of demographic upheaval and concomitant language contact. Tracing the morphological and word-order-based complexity through time in parallel Bible translations, our results show that English is consistently less morphologically complex than its sisters, continuing on its path of morphological simplification. We see a statistically robust trade-off between morphology and word order complexity. We find support for earlier findings in the literature, with the exception of the difference between Dutch and German, which does not transpire in our results.

Double debiased transfer learning for adaptive Huber regression

AbstractThrough exploiting information from the source data to improve the fit performance on the target data, transfer learning estimations for high‐dimensional linear regression models have drawn much attention recently, but few studies focus on statistical inference and robust learning in the presence of heavy‐tailed/asymmetric errors. Using adaptive Huber regression (AHR) to achieve the bias and robustness tradeoff, in this paper we propose a robust transfer learning algorithm with high‐dimensional covariates, then construct valid confidence intervals and hypothesis tests based on the debiased lasso approach. When the transferable sources are known, a two‐step ‐penalized transfer AHR estimator is firstly proposed and the error bounds are established. To correct the biases caused by the lasso penalty, a unified debiasing framework based on the decorrelated score equations is considered to establish asymptotic normality of the debiased lasso transfer AHR estimator. Confidence intervals and hypothesis tests for each component can be constructed. When the transferable sources are unknown, a data‐driven source detection algorithm is proposed with theoretical guarantee. Numerical studies verify the performance of our proposed estimator and confidence intervals, and an application to Genotype‐Tissue Expression data is also presented.

Balancing robustness and adaptation rate for proton therapy of lung cancer patients

IntroductionAn increase in plan robustness leads to a higher dose to organs-at-risk (OARs), and an increased chance of post-treatment toxicities. In contrast, more conformal plans lead to sparing of healthy surrounding tissue at the expense of a higher sensitivity to anatomical changes, requiring costly adaptations. In this study, we assess the trade-off and impact of treatment plan robustness on the adaptation rate. MethodTreatment planning was performed for 40 lung cancer patients, each having a planning 4DCT and up to eight weekly repeated 4DCTs (reCTs). For each patient, plans were made with three different levels of robustness based on setup uncertainty of 3, 6 and 9 mm. These plans were robustly re-evaluated on all reCTs to assess whether the clinical constraints were met. ResultsFor the 3, 6 and 9 mm robustness levels, adaptation rates of 87.5 %, 70.0 % and 57.5 %, respectively, were observed. A mean absolute normal tissue complication probability (NTCP) gain of 2.9 percentage points (pp) was calculated for pneumonitis grade ≥ 2 when transitioning from 9 mm plans to 3 mm plans, 7.6 pp for esophagitis grade ≥ 2, and 2.5 pp for mortality risk 2 years post-treatment. ConclusionThe lowered risk of post treatment toxicities at lower robustness levels is clinically relevant but comes at the expense of more treatment adaptations, particularly in cases where meeting our clinical goals is not compromised by having a dose that is more conformal to the target. The trade-off between workload and reduced NTCP needs to be individually assessed.

Interactive and Multimetric Robustness Tradeoffs in the Colorado River Basin

Interactive and Multimetric Robustness Tradeoffs in the Colorado River Basin

Fourier analysis on robustness of graph convolutional neural networks for skeleton-based action recognition

Using Fourier analysis, we explore the robustness and vulnerability of graph convolutional neural networks (GCNs) for skeleton-based action recognition. We adopt a joint Fourier transform (JFT), a combination of the graph Fourier transform (GFT) and the discrete Fourier transform (DFT), to examine the robustness of adversarially-trained GCNs against adversarial attacks and common corruptions. Experimental results with the NTU RGB+D dataset reveal that adversarial training does not introduce a robustness trade-off between adversarial attacks and low-frequency perturbations, which typically occurs during image classification based on convolutional neural networks. This finding indicates that adversarial training is a practical approach to enhancing robustness against adversarial attacks and common corruptions in skeleton-based action recognition. Furthermore, we find that the Fourier approach cannot explain vulnerability against skeletal part occlusion corruption, which highlights its limitations. These findings extend our understanding of the robustness of GCNs, potentially guiding the development of more robust learning methods for skeleton-based action recognition.

Performance and robustness trade-offs in PIR control of uncertain second-order systems with input disturbances

We investigate the robustness, disturbance attenuation and performance capabilities of a Proportional-Integral-Retarded (PIR) controller when controlling a class of uncertain second-order systems with input disturbances. The purpose of this paper is to provide insight into the trade-offs between these conflicting requirements. This is accomplished by introducing simple delay-dependent stability conditions, expressed in linear matrix inequality form, which are used to establish the trade-offs in PIR control. In light of the obtained stability conditions, we show that improving performance comes at the cost of reduced robustness and disturbance attenuation. To our knowledge, this problem has not been addressed before and these new results provide valuable insights and systematic guidelines for designing these controllers with trade-offs in mind, thereby completing our current understanding of the tuning mechanisms in PIR control. Inspired by the application field of fuel-cell systems, experiments are carried out on a proof-of-concept switching converter prototype to demonstrate the relevance of the obtained results.

A robust optimization model for green supplier selection and order allocation in a closed-loop supply chain considering cap-and-trade mechanism

Due to increasing air pollution, which is a consequence of the environmental effects of production in various industries, green supply chain management (GSCM) has attracted the attention of both scholars and practitioners. Green supplier selection is an important problem in GSCM and seeks to satisfy a firm’s environmental goals as well as its economic targets. In this paper, for the first time, a green supplier selection problem considering both green and non-green evaluation criteria in a closed-loop supply chain is studied, and a cap-and-trade mechanism as a way of controlling the air pollution caused by manufacturers is proposed. To solve this particular problem, we propose a multi-objective robust optimization (RO) model. This specific model is an effective approach to handle uncertainty. A numerical example using randomly generated data, accompanied by subsequent discussion of the proposed approach, is deployed to validate the model. The results prove that the developed model for green supplier selection is able to effectively enhance the decision-making process of the experts. By illustrating the trade-off in robustness between the model and proposed solutions, as well as the effect of the deviation penalty on the closeness of results to the achieved solution, we show how firms can make optimal decisions when assigning the parameters. Furthermore, analyses show that decreasing the allowance amount (cap) and increasing the allowance prices in the cap-and-trade system escalate the firms’ costs, but lower the amount of carbon released. Finally, we show that the cap-and-trade mechanism results in a better solution in terms of the total utility of the supply chain compared to the penalty-based system for a specific range of carbon allowance.

The separation of several organophosphate pesticides on immobilized polysaccharide chiral stationary phases.

While not initially a focus or priority, in recent decades, an emphasis has been placed on the activity of individual enantiomers of widely used pesticides. Of particular note are organophosphorus‐based pesticides like fenamiphos and profenofos, as examples. This work explores the enantioselective high‐performance liquid chromatography (HPLC) separations of seven such organophosphorus pesticides (OP's) on the library of immobilized polysaccharide‐based chiral stationary phases (CSPs) with normal phase hexane/alcohol mixtures. Further exploration of the effect of mobile phase strength and temperature on several of the separations was performed using simple factorial design. Equivalent retention of the first eluting enantiomer of several combinations of temperature and mobile phase was compared for peak shape, selectivity, and resolution. Similarly, equivalent selectivity of several combinations of temperature and mobile phase was compared for peak shape, retention of the first eluting enantiomer, and resolution. The results of this study make available several new chiral separations of the OPs included in the work that were not previously documented, including separations on the three most recently commercialized phases, Chiralpak IH, IJ, and IK. Additionally, sufficient understanding was obtained to be able to predict the trade‐off of resolution, analysis time, peak sharpness (and thus improve limit‐of‐detection [LOD]/limit‐of‐quantification [LOQ]), robustness, and convenience of conditions for further application optimization.

Morphological effects of leading-edge serrations on the acoustic signatures of mixed flow fan

Leading-edge (LE) noise is a common source of broadband noise for fans that can be suppressed using appended LE serrations. We conduct an integrated study of the morphological effects of interval, length, and inclination angle of owl-inspired LE serrations on the aeroacoustic characteristics of a mixed flow fan using experiments, computational fluid dynamics (CFD), and the Ffowcs Williams–Hawkings (FWH) analogy. A novel method for surface noise strength (SNS) visualization was developed based on the FWH analogy with large-eddy simulations to accurately quantify the spatial distributions of acoustic sources. A CFD-informed index is proposed to evaluate the severity of flow separation with the pressure gradient and verified to be effective in examining the chord-wise separation. Acoustic measurements show the robust trade-off solving capability of the serrations under various morphologies, and the SNS visualizations indicate that the separation-induced LE noise is suppressed considerably. One-third octave analyses suggest that extending serration length can lower separation noise more effectively than shrinking the interval over 100–3000 Hz. A smaller interval is more desirable while an optimal length exists in association with tonal noise. Moreover, small inclination angles (≤20°) enable the deceleration of oncoming flows with stagnation relieved, and consequently, further suppress the LE noise, by a flow-buffering effect. Heavy inclination angles (≥40°) induce an additional tip vortex, causing high-coherence turbulence impingement noise and resulting in a drastic increase in broadband noise at frequencies exceeding 4000 Hz. Our study, thus, clarifies the morphological effects of LE serrations on aeroacoustic signatures of rotary devices while providing useful methods for acoustic analyses.

Mutual Adversarial Training: Learning Together is Better Than Going Alone

Recent studies have shown that robustness to adversarial attacks can be transferred across deep neural networks. In other words, we can make a weak model more robust with the help of a strong teacher model. In this paper, we ask if models can &#x201C;learn together&#x201D; and &#x201C;teach each other&#x201D; to achieve better robustness instead of learning from a static teacher.We study how interactions among models affect robustness via knowledge distillation. We propose mutual adversarial training (MAT), in which multiple models are trained together and share the knowledge of adversarial examples to achieve improved robustness. MAT allows robust models to explore a larger space of adversarial samples and find more robust feature spaces and decision boundaries. Through extensive experiments on the CIFAR-10, CIFAR-100, and mini-ImageNet datasets, we demonstrate that MAT can effectively improve model robustness and outperform state-of-the-art methods under white-box attacks. In addition, we show that MAT can also mitigate the robustness trade-off among different perturbation types. Specially, we train specialist models that learn to defend a specific perturbation type and a generalist model that learns to defend multiple perturbation types by learning from the specialists, which brings as much as 13.4% accuracy gain to AT baselines against the union of <i>l</i>&#x221E;, <i>l</i>2, and <i>l</i>1 attacks. Our results show the superiority of the proposed method and demonstrate that collaborative learning is an effective strategy for designing robust models.

Robust capacity planning for sterilisation department of a hospital

ABSTRACT Sterile services departments are special units designed to perform sterilisation operations in an efficient way within a hospital. The delays in sterilisation services cause significant disruptions on surgery schedules and bed management. To prevent the delays, an upper time limit can be imposed on the time spent in the sterilisation services. In this paper, we propose a mathematical modelling approach for the optimum capacity planning of a sterilisation service unit considering the uncertainties in the sterilisation process. The model aims to find the optimum capacity on four tandem steps of the sterilisation whilst at the same time minimising the total cost and keeping the maximum time in the system below a limit. Assuming general distributions for service and interarrival times, an approximation structure based on robust optimisation is used to formulate the maximum time spent in the system. We analysed the structural property of the resulting model and found that the relaxed version of the model is convex. The real data from a large sterilisation services unit is used for computational experiments. The results indicated that the approximation fits well against the simulated maximum time in the system. Other experiments revealed that an upper limit of 7 hours for the sterilisation services balances the cost vs. robustness trade-off.

Goodness-of-fit criteria for hydrological models: Model calibration and performance assessment

This study provides guidelines for the selection of proper goodness-of-fit criteria for the calibration and evaluation of hydrological models. Popular goodness-of-fit criteria and good practices for hydrological modeling are reviewed. The review discusses the advantages and disadvantages of several criteria and is followed by a case study that focuses on the review’s main findings. The main recommendation is for hydrologists to avoid using threshold values to assess model performance and preferably set a proper benchmark series. The case study was developed using the GR5J hydrological model and data from 179 watersheds in the Brazilian Cerrado biome. Several single- and multi-objective functions are used in optimization runs to assess the outcome for different goodness-of-fit criteria. The model performance is evaluated for each optimization run considering overall conditions, i.e., entire time series, and conditions under low- and peak-flow conditions. The study case reinforces that the popular Nash-Sutcliffe efficiency index should be avoided as an objective function. Alternatively, the Kling-Gupta efficiency index showed to be a more reliable criterion, resulting in lower bias for both calibration and validation, and balanced results for both low- and peak-flow conditions. Additionally, combining different criteria in multi-objective functions can result in robust trade-offs. General guidelines are summarized and additional emphasis is given to tropical watersheds where low flows deserve due attention.

Near-Optimal Robust Virtual Controller Placement in 5G Software Defined Networks

Fifth generation (5G) wireless networksare characterized by applying the software defined networking (SDN) and network function virtualization (NFV) concepts to provide higher reliability and scalability, and lower latency for advent data-hungry services. In the initial version of SDN based 5G, a centralized architecture for SDN controller has been proposed, which inherently cannot fulfill those requirements simultaneously. This observation has led us toward distributed architectures, where multiple SDN controllers across the network constitute the control layers. Furthermore, to fully achieve the network flexibility and agility offered by SDN, and also improve resource utilization and consequently to reduce network cost, we incorporate virtualized SDN (vSDN) controllers, enabled through the NFV technology. In vSDN with distributed architecture, controller placement is one of the main challenges, which is an NP-hard combinatorial optimization problem with conflicting objectives: 1) maximizing the reliability of network with low latency; 2) minimizing the total cost of implementation, which can be interpreted as reducing the number of vSDN controllers. In order to achieve such a compromise, in this work, we formulate the problem of vSDN controller placement and introduce some performance metrics to model the latency of communication between controller-switch pairs and also robustness against vSDN controller failures in a network. The introduced formulation is some sorts of submodular optimization, which is exploited to develop several algorithms. Simulations results demonstrate that, with a negligible performance loss, a significant reduction in the number of vSDN controllers is achievable. Finally, latency, cost, and robustness trade-offs are investigated.

The use of TOPSIS-based-desirability function approach to optimize the balances among mechanical performances, energy consumption, and production efficiency of the arc welding process

A couple of models were established to investigate the effects of welding parameters (voltage, wire feed speed, and welding speed) on the resultant mechanical properties of welding bead and welding heat input on the basis of a Box-Behnken experimental design (BBD). The three key mechanical parameters, that is, maximum displacement, peak load, and energy absorption, were processed via the technique in order of preference by similarity to ideal solution (TOPSIS) and Shannon entropy technique. After that, the correlations among the mechanical properties, welding heat input, and three technological variables in the welding process were established herein. Analysis of variance (ANOVA) implied that the heat input relied on voltage, welding speed, wire feed rate, and the interaction effects among these factors. These models act as the basis to achieve the multi-objective optimization problem by the desirability function approach. Results suggest that the welding settings favoring a robust trade-off between minimum welding heat input and maximum mechanical properties involve an intermediate value of wire feed speed, a high value of voltage, and welding speed. This welding parameter combination not only can produce an optimum welding bead with robust mechanical performances but also guarantees the goal of optimum welding heat utilization and production efficiency, in which the high level of welding speed is strongly recommended.

Understanding the Energy vs. Adversarial Robustness Trade-Off in Deep Neural Networks

Understanding the Energy vs. Adversarial Robustness Trade-Off in Deep Neural Networks

Primate differential redoxome (PDR) – A paradigm for understanding neurodegenerative diseases

Despite different phenotypic manifestations, mounting evidence points to similarities in the molecular basis of major neurodegenerative diseases (ND). CNS has evolved to be robust against hazard of ROS, a common perturbation aerobic organisms are confronted with. The trade-off of robustness is system's fragility against rare and unexpected perturbations. Identifying the points of CNS fragility is key for understanding etiology of ND. We postulated that the ‘primate differential redoxome’ (PDR), an assembly of proteins that contain cysteine residues present only in the primate orthologues of mammals, is likely to associate with an added level of regulatory functionalities that enhanced CNS robustness against ROS and facilitated evolution. The PDR contains multiple deterministic and susceptibility factors of major ND, which cluster to form coordinated redox networks regulating various cellular processes. The PDR analysis revealed a potential CNS fragility point, which appears to associates with a non-redundant PINK1-PRKN-SQSTM1(p62) axis coordinating protein homeostasis and mitophagy.