Pooyan Jamshidi

Fatemeh Ghofrani, Mehdi Yaghouti, Pooyan Jamshidi

To advance the understanding of robust deep learning, we delve into the effects of adversarial training on self-supervised and supervised contrastive learning alongside supervised learning. Our analysis uncovers significant disparities between adversarial and clean representations in standard-trained networks across various learning algorithms. Remarkably, adversarial training mitigates these disparities and fosters the convergence of representations toward a universal set, regardless of the learning scheme used. Additionally, increasing the similarity between adversarial and clean representations, particularly near the end of the network, enhances network robustness. These findings offer valuable insights for designing and training effective and robust deep learning networks. Our code is released at \textcolor{magenta}{\url{https://github.com/softsys4ai/CL-Robustness}}.

Md Abir Hossen, Sonam Kharade, Bradley Schmerl et al.

Robotic systems have subsystems with a combinatorially large configuration space and hundreds or thousands of possible software and hardware configuration options interacting non-trivially. The configurable parameters are set to target specific objectives, but they can cause functional faults when incorrectly configured. Finding the root cause of such faults is challenging due to the exponentially large configuration space and the dependencies between the robot's configuration settings and performance. This paper proposes CaRE -- a method for diagnosing the root cause of functional faults through the lens of causality. CaRE abstracts the causal relationships between various configuration options and the robot's performance objectives by learning a causal structure and estimating the causal effects of options on robot performance indicators. We demonstrate CaRE's efficacy by finding the root cause of the observed functional faults and validating the diagnosed root cause by conducting experiments in both physical robots (Husky and Turtlebot 3) and in simulation (Gazebo). Furthermore, we demonstrate that the causal models learned from robots in simulation (e.g., Husky in Gazebo) are transferable to physical robots across different platforms (e.g., Husky and Turtlebot 3).

Mehran Salmani, Saeid Ghafouri, Alireza Sanaee et al.

The use of machine learning (ML) inference for various applications is growing drastically. ML inference services engage with users directly, requiring fast and accurate responses. Moreover, these services face dynamic workloads of requests, imposing changes in their computing resources. Failing to right-size computing resources results in either latency service level objectives (SLOs) violations or wasted computing resources. Adapting to dynamic workloads considering all the pillars of accuracy, latency, and resource cost is challenging. In response to these challenges, we propose InfAdapter, which proactively selects a set of ML model variants with their resource allocations to meet latency SLO while maximizing an objective function composed of accuracy and cost. InfAdapter decreases SLO violation and costs up to 65% and 33%, respectively, compared to a popular industry autoscaler (Kubernetes Vertical Pod Autoscaler).

Ali Mokhtari, Md Abir Hossen, Pooyan Jamshidi et al.

Edge computing enables smart IoT-based systems via concurrent and continuous execution of latency-sensitive machine learning (ML) applications. These edge-based machine learning systems are often battery-powered (i.e., energy-limited). They use heterogeneous resources with diverse computing performance (e.g., CPU, GPU, and/or FPGAs) to fulfill the latency constraints of ML applications. The challenge is to allocate user requests for different ML applications on the Heterogeneous Edge Computing Systems (HEC) with respect to both the energy and latency constraints of these systems. To this end, we study and analyze resource allocation solutions that can increase the on-time task completion rate while considering the energy constraint. Importantly, we investigate edge-friendly (lightweight) multi-objective mapping heuristics that do not become biased toward a particular application type to achieve the objectives; instead, the heuristics consider "fairness" across the concurrent ML applications in their mapping decisions. Performance evaluations demonstrate that the proposed heuristic outperforms widely-used heuristics in heterogeneous systems in terms of the latency and energy objectives, particularly, at low to moderate request arrival rates. We observed 8.9% improvement in on-time task completion rate and 12.6% in energy-saving without imposing any significant overhead on the edge system.

Saeid Ghafouri, Kamran Razavi, Mehran Salmani et al.

Efficiently optimizing multi-model inference pipelines for fast, accurate, and cost-effective inference is a crucial challenge in machine learning production systems, given their tight end-to-end latency requirements. To simplify the exploration of the vast and intricate trade-off space of latency, accuracy, and cost in inference pipelines, providers frequently opt to consider one of them. However, the challenge lies in reconciling latency, accuracy, and cost trade-offs. To address this challenge and propose a solution to efficiently manage model variants in inference pipelines, we present IPA, an online deep learning Inference Pipeline Adaptation system that efficiently leverages model variants for each deep learning task. Model variants are different versions of pre-trained models for the same deep learning task with variations in resource requirements, latency, and accuracy. IPA dynamically configures batch size, replication, and model variants to optimize accuracy, minimize costs, and meet user-defined latency Service Level Agreements (SLAs) using Integer Programming. It supports multi-objective settings for achieving different trade-offs between accuracy and cost objectives while remaining adaptable to varying workloads and dynamic traffic patterns. Navigating a wider variety of configurations allows \namex{} to achieve better trade-offs between cost and accuracy objectives compared to existing methods. Extensive experiments in a Kubernetes implementation with five real-world inference pipelines demonstrate that IPA improves end-to-end accuracy by up to 21% with a minimal cost increase. The code and data for replications are available at https://github.com/reconfigurable-ml-pipeline/ipa.

Hamed Damirchi, Forest Agostinelli, Pooyan Jamshidi

Monolithic neural networks that make use of a single set of weights to learn useful representations for downstream tasks explicitly dismiss the compositional nature of data generation processes. This characteristic exists in data where every instance can be regarded as the combination of an identity concept, such as the shape of an object, combined with modifying concepts, such as orientation, color, and size. The dismissal of compositionality is especially detrimental in robotics, where state estimation relies heavily on the compositional nature of physical mechanisms (e.g., rotations and transformations) to model interactions. To accommodate this data characteristic, modular networks have been proposed. However, a lack of structure in each module's role, and modular network-specific issues such as module collapse have restricted their usability. We propose a modular network architecture that accommodates the mentioned decompositional concept by proposing a unique structure that splits the modules into predetermined roles. Additionally, we provide regularizations that improve the resiliency of the modular network to the problem of module collapse while improving the decomposition accuracy of the model.

Mohammadamin Abedi, Yanni Iouannou, Pooyan Jamshidi et al.

Deep Neural Networks (DNNs) have become an essential component in many application domains including web-based services. A variety of these services require high throughput and (close to) real-time features, for instance, to respond or react to users' requests or to process a stream of incoming data on time. However, the trend in DNN design is toward larger models with many layers and parameters to achieve more accurate results. Although these models are often pre-trained, the computational complexity in such large models can still be relatively significant, hindering low inference latency. Implementing a caching mechanism is a typical systems engineering solution for speeding up a service response time. However, traditional caching is often not suitable for DNN-based services. In this paper, we propose an end-to-end automated solution to improve the performance of DNN-based services in terms of their computational complexity and inference latency. Our caching method adopts the ideas of self-distillation of DNN models and early exits. The proposed solution is an automated online layer caching mechanism that allows early exiting of a large model during inference time if the cache model in one of the early exits is confident enough for final prediction. One of the main contributions of this paper is that we have implemented the idea as an online caching, meaning that the cache models do not need access to training data and perform solely based on the incoming data at run-time, making it suitable for applications using pre-trained models. Our experiments results on two downstream tasks (face and object classification) show that, on average, caching can reduce the computational complexity of those services up to 58\% (in terms of FLOPs count) and improve their inference latency up to 46\% with low to zero reduction in accuracy.

Fatemeh Ghofrani, Pooyan Jamshidi

Self-supervised learning (SSL) has significantly advanced image representation learning, yet efficiency challenges persist, particularly with adversarial training. Many SSL methods require extensive epochs to achieve convergence, a demand further amplified in adversarial settings. To address this inefficiency, we revisit the robust EMP-SSL framework, emphasizing the importance of increasing the number of crops per image to accelerate learning. Unlike traditional contrastive learning, robust EMP-SSL leverages multi-crop sampling, integrates an invariance term and regularization, and reduces training epochs, enhancing time efficiency. Evaluated with both standard linear classifiers and multi-patch embedding aggregation, robust EMP-SSL provides new insights into SSL evaluation strategies. Our results show that robust crop-based EMP-SSL not only accelerates convergence but also achieves a superior balance between clean accuracy and adversarial robustness, outperforming multi-crop embedding aggregation. Additionally, we extend this approach with free adversarial training in Multi-Crop SSL, introducing the Cost-Free Adversarial Multi-Crop Self-Supervised Learning (CF-AMC-SSL) method. CF-AMC-SSL demonstrates the effectiveness of free adversarial training in reducing training time while simultaneously improving clean accuracy and adversarial robustness. These findings underscore the potential of CF-AMC-SSL for practical SSL applications. Our code is publicly available at https://github.com/softsys4ai/CF-AMC-SSL.

Miguel Velez, Pooyan Jamshidi, Norbert Siegmund et al.

Performance-influence models can help stakeholders understand how and where configuration options and their interactions influence the performance of a system. With this understanding, stakeholders can debug performance behavior and make deliberate configuration decisions. Current black-box techniques to build such models combine various sampling and learning strategies, resulting in tradeoffs between measurement effort, accuracy, and interpretability. We present Comprex, a white-box approach to build performance-influence models for configurable systems, combining insights of local measurements, dynamic taint analysis to track options in the implementation, compositionality, and compression of the configuration space, without relying on machine learning to extrapolate incomplete samples. Our evaluation on 4 widely-used, open-source projects demonstrates that Comprex builds similarly accurate performance-influence models to the most accurate and expensive black-box approach, but at a reduced cost and with additional benefits from interpretable and local models.

Md Abir Hossen, Sonam Kharade, Jason M. O'Kane et al.

Robotic systems are typically composed of various subsystems, such as localization and navigation, each encompassing numerous configurable components (e.g., selecting different planning algorithms). Once an algorithm has been selected for a component, its associated configuration options must be set to the appropriate values. Configuration options across the system stack interact non-trivially. Finding optimal configurations for highly configurable robots to achieve desired performance poses a significant challenge due to the interactions between configuration options across software and hardware that result in an exponentially large and complex configuration space. These challenges are further compounded by the need for transferability between different environments and robotic platforms. Data efficient optimization algorithms (e.g., Bayesian optimization) have been increasingly employed to automate the tuning of configurable parameters in cyber-physical systems. However, such optimization algorithms converge at later stages, often after exhausting the allocated budget (e.g., optimization steps, allotted time) and lacking transferability. This paper proposes CURE -- a method that identifies causally relevant configuration options, enabling the optimization process to operate in a reduced search space, thereby enabling faster optimization of robot performance. CURE abstracts the causal relationships between various configuration options and robot performance objectives by learning a causal model in the source (a low-cost environment such as the Gazebo simulator) and applying the learned knowledge to perform optimization in the target (e.g., Turtlebot 3 physical robot). We demonstrate the effectiveness and transferability of CURE by conducting experiments that involve varying degrees of deployment changes in both physical robots and simulation.

Omid Gheibi, Christian Kästner, Pooyan Jamshidi

Performance-influence models are beneficial for understanding how configurations affect system performance, but their creation is challenging due to the exponential growth of configuration spaces. While gray-box approaches leverage selective "structural knowledge" (like the module execution graph of the system) to improve modeling, the relationship between this knowledge, a system's characteristics (we call them "structural aspects"), and potential model improvements is not well understood. This paper addresses this gap by formally investigating how variations in structural aspects (e.g., the number of modules and options per module) and the level of structural knowledge impact the creation of "opportunities" for improved "modular performance modeling". We introduce and quantify the concept of modeling "hardness", defined as the inherent difficulty of performance modeling. Through controlled experiments with synthetic system models, we establish an "analytical matrix" to measure these concepts. Our findings show that modeling hardness is primarily driven by the number of modules and configuration options per module. More importantly, we demonstrate that both higher levels of structural knowledge and increased modeling hardness significantly enhance the opportunity for improvement. The impact of these factors varies by performance metric; for ranking accuracy (e.g., in debugging task), structural knowledge is more dominant, while for prediction accuracy (e.g., in resource management task), hardness plays a stronger role. These results provide actionable insights for system designers, guiding them to strategically allocate time and select appropriate modeling approaches based on a system's characteristics and a given task's objectives.

Rahul Goel, Modar Sulaiman, Kimia Noorbakhsh et al.

Facial images disclose many hidden personal traits such as age, gender, race, health, emotion, and psychology. Understanding these traits will help to classify the people in different attributes. In this paper, we have presented a novel method for classifying images using a pretrained transformer model. We apply the pretrained transformer for the binary classification of facial images in criminal and non-criminal classes. The pretrained transformer of GPT-2 is trained to generate text and then fine-tuned to classify facial images. During the finetuning process with images, most of the layers of GT-2 are frozen during backpropagation and the model is frozen pretrained transformer (FPT). The FPT acts as a universal image classifier, and this paper shows the application of FPT on facial images. We also use our FPT on encrypted images for classification. Our FPT shows high accuracy on both raw facial images and encrypted images. We hypothesize the meta-learning capacity FPT gained because of its large size and trained on a large size with theory and experiments. The GPT-2 trained to generate a single word token at a time, through the autoregressive process, forced to heavy-tail distribution. Then the FPT uses the heavy-tail property as its meta-learning capacity for classifying images. Our work shows one way to avoid bias during the machine classification of images.The FPT encodes worldly knowledge because of the pretraining of one text, which it uses during the classification. The statistical error of classification is reduced because of the added context gained from the text.Our paper shows the ethical dimension of using encrypted data for classification.Criminal images are sensitive to share across the boundary but encrypted largely evades ethical concern.FPT showing good classification accuracy on encrypted images shows promise for further research on privacy-preserving machine learning.

Md Shahriar Iqbal, Rahul Krishna, Mohammad Ali Javidian et al.

Modern computer systems are highly configurable, with the total variability space sometimes larger than the number of atoms in the universe. Understanding and reasoning about the performance behavior of highly configurable systems, over a vast and variable space, is challenging. State-of-the-art methods for performance modeling and analyses rely on predictive machine learning models, therefore, they become (i) unreliable in unseen environments (e.g., different hardware, workloads), and (ii) may produce incorrect explanations. To tackle this, we propose a new method, called Unicorn, which (i) captures intricate interactions between configuration options across the software-hardware stack and (ii) describes how such interactions can impact performance variations via causal inference. We evaluated Unicorn on six highly configurable systems, including three on-device machine learning systems, a video encoder, a database management system, and a data analytics pipeline. The experimental results indicate that Unicorn outperforms state-of-the-art performance debugging and optimization methods in finding effective repairs for performance faults and finding configurations with near-optimal performance. Further, unlike the existing methods, the learned causal performance models reliably predict performance for new environments.

Kimia Noorbakhsh, Modar Sulaiman, Mahdi Sharifi et al.

Solving symbolic mathematics has always been of in the arena of human ingenuity that needs compositional reasoning and recurrence. However, recent studies have shown that large-scale language models such as transformers are universal and surprisingly can be trained as a sequence-to-sequence task to solve complex mathematical equations. These large transformer models need humongous amounts of training data to generalize to unseen symbolic mathematics problems. In this paper, we present a sample efficient way of solving the symbolic tasks by first pretraining the transformer model with language translation and then fine-tuning the pretrained transformer model to solve the downstream task of symbolic mathematics. We achieve comparable accuracy on the integration task with our pretrained model while using around $1.5$ orders of magnitude less number of training samples with respect to the state-of-the-art deep learning for symbolic mathematics. The test accuracy on differential equation tasks is considerably lower comparing with integration as they need higher order recursions that are not present in language translations. We propose the generalizability of our pretrained language model from Anna Karenina Principle (AKP). We pretrain our model with different pairs of language translations. Our results show language bias in solving symbolic mathematics tasks. Finally, we study the robustness of the fine-tuned model on symbolic math tasks against distribution shift, and our approach generalizes better in distribution shift scenarios for the function integration.

Mohammad Ali Javidian, Om Pandey, Pooyan Jamshidi

One of the most critical problems in transfer learning is the task of domain adaptation, where the goal is to apply an algorithm trained in one or more source domains to a different (but related) target domain. This paper deals with domain adaptation in the presence of covariate shift while invariances exist across domains. One of the main limitations of existing causal inference methods for solving this problem is scalability. To overcome this difficulty, we propose SCTL, an algorithm that avoids an exhaustive search and identifies invariant causal features across source and target domains based on Markov blanket discovery. SCTL does not require having prior knowledge of the causal structure, the type of interventions, or the intervention targets. There is an intrinsic locality associated with SCTL that makes it practically scalable and robust because local causal discovery increases the power of computational independence tests and makes the task of domain adaptation computationally tractable. We show the scalability and robustness of SCTL for domain adaptation using synthetic and real data sets in low-dimensional and high-dimensional settings.

Md. Musfiqur Rahman, Ayman Rasheed, Md. Mosaddek Khan et al.

Causal structure discovery from observational data is fundamental to the causal understanding of autonomous systems such as medical decision support systems, advertising campaigns and self-driving cars. This is essential to solve well-known causal decision making and prediction problems associated with those real-world applications. Recently, recursive causal discovery algorithms have gained particular attention among the research community due to their ability to provide good results by using Conditional Independent (CI) tests in smaller sub-problems. However, each of such algorithms needs a refinement function to remove undesired causal relations of the discovered graphs. Notably, with the increase of the problem size, the computation cost (i.e., the number of CI-tests) of the refinement function makes an algorithm expensive to deploy in practice. This paper proposes a generic causal structure refinement strategy that can locate the undesired relations with a small number of CI-tests, thus speeding up the algorithm for large and complex problems. We theoretically prove the correctness of our algorithm. We then empirically evaluate its performance against the state-of-the-art algorithms in terms of solution quality and completion time in synthetic and real datasets.

Rahul Krishna, Md Shahriar Iqbal, Mohammad Ali Javidian et al.

Modern computing platforms are highly-configurable with thousands of interacting configurations. However, configuring these systems is challenging. Erroneous configurations can cause unexpected non-functional faults. This paper proposes CADET (short for Causal Debugging Toolkit) that enables users to identify, explain, and fix the root cause of non-functional faults early and in a principled fashion. CADET builds a causal model by observing the performance of the system under different configurations. Then, it uses casual path extraction followed by counterfactual reasoning over the causal model to: (a) identify the root causes of non-functional faults, (b) estimate the effects of various configurable parameters on the performance objective(s), and (c) prescribe candidate repairs to the relevant configuration options to fix the non-functional fault. We evaluated CADET on 5 highly-configurable systems deployed on 3 NVIDIA Jetson systems-on-chip. We compare CADET with state-of-the-art configuration optimization and ML-based debugging approaches. The experimental results indicate that CADET can find effective repairs for faults in multiple non-functional properties with (at most) 17% more accuracy, 28% higher gain, and $40\times$ speed-up than other ML-based performance debugging methods. Compared to multi-objective optimization approaches, CADET can find fixes (at most) $9\times$ faster with comparable or better performance gain. Our case study of non-functional faults reported in NVIDIA's forum show that CADET can find $14%$ better repairs than the experts' advice in less than 30 minutes.

Mohammad Ali Javidian, Marco Valtorta, Pooyan Jamshidi

This paper provides a graphical characterization of Markov blankets in chain graphs (CGs) under the Lauritzen-Wermuth-Frydenberg (LWF) interpretation. The characterization is different from the well-known one for Bayesian networks and generalizes it. We provide a novel scalable and sound algorithm for Markov blanket discovery in LWF CGs and prove that the Grow-Shrink algorithm, the IAMB algorithm, and its variants are still correct for Markov blanket discovery in LWF CGs under the same assumptions as for Bayesian networks. We provide a sound and scalable constraint-based framework for learning the structure of LWF CGs from faithful causally sufficient data and prove its correctness when the Markov blanket discovery algorithms in this paper are used. Our proposed algorithms compare positively/competitively against the state-of-the-art LCD (Learn Chain graphs via Decomposition) algorithm, depending on the algorithm that is used for Markov blanket discovery. Our proposed algorithms make a broad range of inference/learning problems computationally tractable and more reliable because they exploit locality.

Mohammad Ali Javidian, Marco Valtorta, Pooyan Jamshidi

LWF chain graphs combine directed acyclic graphs and undirected graphs. We present a PC-like algorithm that finds the structure of chain graphs under the faithfulness assumption to resolve the problem of scalability of the proposed algorithm by Studeny (1997). We prove that our PC-like algorithm is order dependent, in the sense that the output can depend on the order in which the variables are given. This order dependence can be very pronounced in high-dimensional settings. We propose two modifications of the PC-like algorithm that remove part or all of this order dependence. Simulation results under a variety of settings demonstrate the competitive performance of the PC-like algorithms in comparison with the decomposition-based method, called LCD algorithm, proposed by Ma et al. (2008) in low-dimensional settings and improved performance in high-dimensional settings.

Yang Ren, Gregory Gay, Christian Kästner et al.

Modern systems are built using development frameworks. These frameworks have a major impact on how the resulting system executes, how configurations are managed, how it is tested, and how and where it is deployed. Machine learning (ML) frameworks and the systems developed using them differ greatly from traditional frameworks. Naturally, the issues that manifest in such frameworks may differ as well---as may the behavior of developers addressing those issues. We are interested in characterizing the system-related issues---issues impacting performance, memory and resource usage, and other quality attributes---that emerge in ML frameworks, and how they differ from those in traditional frameworks. We have conducted a moderate-scale exploratory study analyzing real-world system-related issues from 10 popular machine learning frameworks. Our findings offer implications for the development of machine learning systems, including differences in the frequency of occurrence of certain issue types, observations regarding the impact of debate and time on issue correction, and differences in the specialization of developers. We hope that this exploratory study will enable developers to improve their expectations, plan for risk, and allocate resources accordingly when making use of the tools provided by these frameworks to develop ML-based systems.

Mahdi Fahmideh, Mohsen Sharifi, Pooyan Jamshidi

Service-orientation is a promising paradigm that enables the engineering of large-scale distributed software systems using rigorous software development processes. The existing problem is that every service-oriented software development project often requires a customized development process that provides specific service-oriented software engineering tasks in support of requirements unique to that project. To resolve this problem and allow situational method engineering, we have defined a set of method fragments in support of the engineering of the project-specific service-oriented software development processes. We have derived the proposed method fragments from the recurring features of eleven prominent service-oriented software development methodologies using a systematic mining approach. We have added these new fragments to the repository of OPEN Process Framework to make them available to software engineers as reusable fragments using this well-known method repository. Keyword. Service-Oriented Software Development, OPEN Process Framework, OPF Repository, Method Fragment, Situational Method Engineering

Mohammad Ali Javidian, Marco Valtorta, Pooyan Jamshidi

We address the problem of finding a minimal separator in an Andersson-Madigan-Perlman chain graph (AMP CG), namely, finding a set Z of nodes that separates a given nonadjacent pair of nodes such that no proper subset of Z separates that pair. We analyze several versions of this problem and offer polynomial-time algorithms for each. These include finding a minimal separator from a restricted set of nodes, finding a minimal separator for two given disjoint sets, and testing whether a given separator is minimal. To address the problem of learning the structure of AMP CGs from data, we show that the PC-like algorithm (Pena, 2012) is order-dependent, in the sense that the output can depend on the order in which the variables are given. We propose several modifications of the PC-like algorithm that remove part or all of this order-dependence. We also extend the decomposition-based approach for learning Bayesian networks (BNs) proposed by (Xie et al., 2006) to learn AMP CGs, which include BNs as a special case, under the faithfulness assumption. We prove the correctness of our extension using the minimal separator results. Using standard benchmarks and synthetically generated models and data in our experiments demonstrate the competitive performance of our decomposition-based method, called LCD-AMP, in comparison with the (modified versions of) PC-like algorithm. The LCD-AMP algorithm usually outperforms the PC-like algorithm, and our modifications of the PC-like algorithm learn structures that are more similar to the underlying ground truth graphs than the original PC-like algorithm, especially in high-dimensional settings. In particular, we empirically show that the results of both algorithms are more accurate and stabler when the sample size is reasonably large and the underlying graph is sparse.

Md Shahriar Iqbal, Jianhai Su, Lars Kotthoff et al.

The design of machine learning systems often requires trading off different objectives, for example, prediction error and energy consumption for deep neural networks (DNNs). Typically, no single design performs well in all objectives; therefore, finding Pareto-optimal designs is of interest. The search for Pareto-optimal designs involves evaluating designs in an iterative process, and the measurements are used to evaluate an acquisition function that guides the search process. However, measuring different objectives incurs different costs. For example, the cost of measuring the prediction error of DNNs is orders of magnitude higher than that of measuring the energy consumption of a pre-trained DNN, as it requires re-training the DNN. Current state-of-the-art methods do not consider this difference in objective evaluation cost, potentially incurring expensive evaluations of objective functions in the optimization process. In this paper, we develop a novel decoupled and cost-aware multi-objective optimization algorithm, we call Flexible Multi-Objective Bayesian Optimization (FlexiBO) to address this issue. FlexiBO weights the improvement of the hypervolume of the Pareto region by the measurement cost of each objective to balance the expense of collecting new information with the knowledge gained through objective evaluations, preventing us from performing expensive measurements for little to no gain. We evaluate FlexiBO on seven state-of-the-art DNNs for image recognition, natural language processing (NLP), and speech-to-text translation. Our results indicate that, given the same total experimental budget, FlexiBO discovers designs with 4.8$\%$ to 12.4$\%$ lower hypervolume error than the best method in state-of-the-art multi-objective optimization.

Ying Meng, Jianhai Su, Jason O'Kane et al.

There has been extensive research on developing defense techniques against adversarial attacks; however, they have been mainly designed for specific model families or application domains, therefore, they cannot be easily extended. Based on the design philosophy of ensemble of diverse weak defenses, we propose ATHENA---a flexible and extensible framework for building generic yet effective defenses against adversarial attacks. We have conducted a comprehensive empirical study to evaluate several realizations of ATHENA with four threat models including zero-knowledge, black-box, gray-box, and white-box. We also explain (i) why diversity matters, (ii) the generality of the defense framework, and (iii) the overhead costs incurred by ATHENA.

Rahul Krishna, Vivek Nair, Pooyan Jamshidi et al.

As software systems grow in complexity and the space of possible configurations increases exponentially, finding the near-optimal configuration of a software system becomes challenging. Recent approaches address this challenge by learning performance models based on a sample set of configurations. However, collecting enough sample configurations can be very expensive since each such sample requires configuring, compiling, and executing the entire system using a complex test suite. When learning on new data is too expensive, it is possible to use \textit{Transfer Learning} to "transfer" old lessons to the new context. Traditional transfer learning has a number of challenges, specifically, (a) learning from excessive data takes excessive time, and (b) the performance of the models built via transfer can deteriorate as a result of learning from a poor source. To resolve these problems, we propose a novel transfer learning framework called BEETLE, which is a "bellwether"-based transfer learner that focuses on identifying and learning from the most relevant source from amongst the old data. This paper evaluates BEETLE with 57 different software configuration problems based on five software systems (a video encoder, an SAT solver, a SQL database, a high-performance C-compiler, and a streaming data analytics tool). In each of these cases, BEETLE found configurations that are as good as or better than those found by other state-of-the-art transfer learners while requiring only a fraction ($\frac{1}{7}$th) of the measurements needed by those other methods. Based on these results, we say that BEETLE is a new high-water mark in optimally configuring software.

Markos Viggiato, Johnatan Oliveira, Eduardo Figueiredo et al.

Code changes are performed differently in the mobile and non-mobile platforms. Prior work has investigated the differences in specific platforms. However, we still lack a deeper understanding of how code changes evolve across different software platforms. In this paper, we present a study aiming at investigating the frequency of changes and how source code, build and test changes co-evolve in mobile and non-mobile platforms. We developed regression models to explain which factors influence the frequency of changes and applied the Apriori algorithm to find types of changes that frequently co-occur. Our findings show that non-mobile repositories have a higher number of commits per month and our regression models suggest that being mobile significantly impacts on the number of commits in a negative direction when controlling for confound factors, such as code size. We also found that developers do not usually change source code files together with build or test files. We argue that our results can provide valuable information for developers on how changes are performed in different platforms so that practices adopted in successful software systems can be followed.

Ahmad Banijamali, Pooyan Jamshidi, Pasi Kuvaja et al.

Connecting vehicles to cloud platforms has enabled innovative business scenarios while raising new quality concerns, such as reliability and scalability, which must be addressed by research. Cloud-native architectures based on microservices are a recent approach to enable continuous delivery and to improve service reliability and scalability. We propose an approach for restructuring cloud platform architectures in the automotive domain into a microservices architecture. To this end, we adopted and implemented microservices patterns from literature to design the cloud-native automotive architecture and conducted a laboratory experiment to evaluate the reliability and scalability of microservices in the context of a real-world project in the automotive domain called Eclipse Kuksa. Findings indicated that the proposed architecture could handle the continuous software delivery over-the-air by sending automatic control messages to a vehicular setting. Different patterns enabled us to make changes or interrupt services without extending the impact to others. The results of this study provide evidences that microservices are a potential design solution when dealing with service failures and high payload on cloud-based services in the automotive domain.

Nabor C. Mendonca, Pooyan Jamshidi, David Garlan et al.

A self-adaptive system can dynamically monitor and adapt its behavior to preserve or enhance its quality attributes under uncertain operating conditions. This article identifies key challenges for the development of microservice applications as self-adaptive systems, using a cloud-based intelligent video surveillance application as a motivating example. It also suggests potential new directions for addressing most of the identified challenges by leveraging existing microservice practices and technologies.

Mohammad Ali Javidian, Marco Valtorta, Pooyan Jamshidi

This paper deals with multivariate regression chain graphs (MVR CGs), which were introduced by Cox and Wermuth [3,4] to represent linear causal models with correlated errors. We consider the PC-like algorithm for structure learning of MVR CGs, which is a constraint-based method proposed by Sonntag and Peña in [18]. We show that the PC-like algorithm is order-dependent, in the sense that the output can depend on the order in which the variables are given. This order-dependence is a minor issue in low-dimensional settings. However, it can be very pronounced in high-dimensional settings, where it can lead to highly variable results. We propose two modifications of the PC-like algorithm that remove part or all of this order-dependence. Simulations under a variety of settings demonstrate the competitive performance of our algorithms in comparison with the original PC-like algorithm in low-dimensional settings and improved performance in high-dimensional settings.

Aaron M. Roth, Nicholay Topin, Pooyan Jamshidi et al.

There is a growing desire in the field of reinforcement learning (and machine learning in general) to move from black-box models toward more "interpretable AI." We improve interpretability of reinforcement learning by increasing the utility of decision tree policies learned via reinforcement learning. These policies consist of a decision tree over the state space, which requires fewer parameters to express than traditional policy representations. Existing methods for creating decision tree policies via reinforcement learning focus on accurately representing an action-value function during training, but this leads to much larger trees than would otherwise be required. To address this shortcoming, we propose a novel algorithm which only increases tree size when the estimated discounted future reward of the overall policy would increase by a sufficient amount. Through evaluation in a simulated environment, we show that its performance is comparable or superior to traditional tree-based approaches and that it yields a more succinct policy. Additionally, we discuss tuning parameters to control the tradeoff between optimizing for smaller tree size or for overall reward.

Miguel Velez, Pooyan Jamshidi, Florian Sattler et al.

Stakeholders of configurable systems are often interested in knowing how configuration options influence the performance of a system to facilitate, for example, the debugging and optimization processes of these systems. Several black-box approaches can be used to obtain this information, but they either sample a large number of configurations to make accurate predictions or miss important performance-influencing interactions when sampling few configurations. Furthermore, black-box approaches cannot pinpoint the parts of a system that are responsible for performance differences among configurations. This article proposes ConfigCrusher, a white-box performance analysis that inspects the implementation of a system to guide the performance analysis, exploiting several insights of configurable systems in the process. ConfigCrusher employs a static data-flow analysis to identify how configuration options may influence control-flow statements and instruments code regions, corresponding to these statements, to dynamically analyze the influence of configuration options on the regions' performance. Our evaluation on 10 configurable systems shows the feasibility of our white-box approach to more efficiently build performance-influence models that are similar to or more accurate than current state of the art approaches. Overall, we showcase the benefits of white-box performance analyses and their potential to outperform black-box approaches and provide additional information for analyzing configurable systems.

Md Shahriar Iqbal, Lars Kotthoff, Pooyan Jamshidi

Modern deep neural network (DNN) systems are highly configurable with large a number of options that significantly affect their non-functional behavior, for example inference time and energy consumption. Performance models allow to understand and predict the effects of such configuration options on system behavior, but are costly to build because of large configuration spaces. Performance models from one environment cannot be transferred directly to another; usually models are rebuilt from scratch for different environments, for example different hardware. Recently, transfer learning methods have been applied to reuse knowledge from performance models trained in one environment in another. In this paper, we perform an empirical study to understand the effectiveness of different transfer learning strategies for building performance models of DNN systems. Our results show that transferring information on the most influential configuration options and their interactions is an effective way of reducing the cost to build performance models in new environments.

Pooyan Jamshidi, Javier Cámara, Bradley Schmerl et al.

Modern cyber-physical systems (e.g., robotics systems) are typically composed of physical and software components, the characteristics of which are likely to change over time. Assumptions about parts of the system made at design time may not hold at run time, especially when a system is deployed for long periods (e.g., over decades). Self-adaptation is designed to find reconfigurations of systems to handle such run-time inconsistencies. Planners can be used to find and enact optimal reconfigurations in such an evolving context. However, for systems that are highly configurable, such planning becomes intractable due to the size of the adaptation space. To overcome this challenge, in this paper we explore an approach that (a) uses machine learning to find Pareto-optimal configurations without needing to explore every configuration and (b) restricts the search space to such configurations to make planning tractable. We explore this in the context of robot missions that need to consider task timeliness and energy consumption. An independent evaluation shows that our approach results in high-quality adaptation plans in uncertain and adversarial environments.

Mohammad Ali Javidian, Pooyan Jamshidi, Marco Valtorta

Modern systems (e.g., deep neural networks, big data analytics, and compilers) are highly configurable, which means they expose different performance behavior under different configurations. The fundamental challenge is that one cannot simply measure all configurations due to the sheer size of the configuration space. Transfer learning has been used to reduce the measurement efforts by transferring knowledge about performance behavior of systems across environments. Previously, research has shown that statistical models are indeed transferable across environments. In this work, we investigate identifiability and transportability of causal effects and statistical relations in highly-configurable systems. Our causal analysis agrees with previous exploratory analysis \cite{Jamshidi17} and confirms that the causal effects of configuration options can be carried over across environments with high confidence. We expect that the ability to carry over causal relations will enable effective performance analysis of highly-configurable systems.

Cor-Paul Bezemer, Simon Eismann, Vincenzo Ferme et al.

DevOps is a modern software engineering paradigm that is gaining widespread adoption in industry. The goal of DevOps is to bring software changes into production with a high frequency and fast feedback cycles. This conflicts with software quality assurance activities, particularly with respect to performance. For instance, performance evaluation activities -- such as load testing -- require a considerable amount of time to get statistically significant results. We conducted an industrial survey to get insights into how performance is addressed in industrial DevOps settings. In particular, we were interested in the frequency of executing performance evaluations, the tools being used, the granularity of the obtained performance data, and the use of model-based techniques. The survey responses, which come from a wide variety of participants from different industry sectors, indicate that the complexity of performance engineering approaches and tools is a barrier for wide-spread adoption of performance analysis in DevOps. The implication of our results is that performance analysis tools need to have a short learning curve, and should be easy to integrate into the DevOps pipeline.

Vivek Nair, Rahul Krishna, Tim Menzies et al.

As software systems grow in complexity, the space of possible configurations grows exponentially. Within this increasing complexity, developers, maintainers, and users cannot keep track of the interactions between all the various configuration options. Finding the optimally performing configuration of a software system for a given setting is challenging. Recent approaches address this challenge by learning performance models based on a sample set of configurations. However, collecting enough data on enough sample configurations can be very expensive since each such sample requires configuring, compiling and executing the entire system against a complex test suite. The central insight of this paper is that choosing a suitable source (a.k.a. "bellwether") to learn from, plus a simple transfer learning scheme will often outperform much more complex transfer learning methods. Using this insight, this paper proposes BEETLE, a novel bellwether based transfer learning scheme, which can identify a suitable source and use it to find near-optimal configurations of a software system. BEETLE significantly reduces the cost (in terms of the number of measurements of sample configuration) to build performance models. We evaluate our approach with 61 scenarios based on 5 software systems and demonstrate that BEETLE is beneficial in all cases. This approach offers a new highwater mark in configuring software systems. Specifically, BEETLE can find configurations that are as good or better as those found by anything else while requiring only 1/7th of the evaluations needed by the state-of-the-art.

Aldeida Aleti, Catia Trubiani, André van Hoorn et al.

Software engineers often have to estimate the performance of a software system before having full knowledge of the system parameters, such as workload and operational profile. These uncertain parameters inevitably affect the accuracy of quality evaluations, and the ability to judge if the system can continue to fulfil performance requirements if parameter results are different from expected. Previous work has addressed this problem by modelling the potential values of uncertain parameters as probability distribution functions, and estimating the robustness of the system using Monte Carlo-based methods. These approaches require a large number of samples, which results in high computational cost and long waiting times. To address the computational inefficiency of existing approaches, we employ Polynomial Chaos Expansion (PCE) as a rigorous method for uncertainty propagation and further extend its use to robust performance estimation. The aim is to assess if the software system is robust, i.e., it can withstand possible changes in parameter values, and continue to meet performance requirements. PCE is a very efficient technique, and requires significantly less computations to accurately estimate the distribution of performance indices. Through three very different case studies from different phases of software development and heterogeneous application domains, we show that PCE can accurately (>97\%) estimate the robustness of various performance indices, and saves up to 225 hours of performance evaluation time when compared to Monte Carlo Simulation.

Pooyan Jamshidi, Norbert Siegmund, Miguel Velez et al.

Modern software systems provide many configuration options which significantly influence their non-functional properties. To understand and predict the effect of configuration options, several sampling and learning strategies have been proposed, albeit often with significant cost to cover the highly dimensional configuration space. Recently, transfer learning has been applied to reduce the effort of constructing performance models by transferring knowledge about performance behavior across environments. While this line of research is promising to learn more accurate models at a lower cost, it is unclear why and when transfer learning works for performance modeling. To shed light on when it is beneficial to apply transfer learning, we conducted an empirical study on four popular software systems, varying software configurations and environmental conditions, such as hardware, workload, and software versions, to identify the key knowledge pieces that can be exploited for transfer learning. Our results show that in small environmental changes (e.g., homogeneous workload change), by applying a linear transformation to the performance model, we can understand the performance behavior of the target environment, while for severe environmental changes (e.g., drastic workload change) we can transfer only knowledge that makes sampling more efficient, e.g., by reducing the dimensionality of the configuration space.

Hamid Arabnejad, Claus Pahl, Pooyan Jamshidi et al.

A goal of cloud service management is to design self-adaptable auto-scaler to react to workload fluctuations and changing the resources assigned. The key problem is how and when to add/remove resources in order to meet agreed service-level agreements. Reducing application cost and guaranteeing service-level agreements (SLAs) are two critical factors of dynamic controller design. In this paper, we compare two dynamic learning strategies based on a fuzzy logic system, which learns and modifies fuzzy scaling rules at runtime. A self-adaptive fuzzy logic controller is combined with two reinforcement learning (RL) approaches: (i) Fuzzy SARSA learning (FSL) and (ii) Fuzzy Q-learning (FQL). As an off-policy approach, Q-learning learns independent of the policy currently followed, whereas SARSA as an on-policy always incorporates the actual agent's behavior and leads to faster learning. Both approaches are implemented and compared in their advantages and disadvantages, here in the OpenStack cloud platform. We demonstrate that both auto-scaling approaches can handle various load traffic situations, sudden and periodic, and delivering resources on demand while reducing operating costs and preventing SLA violations. The experimental results demonstrate that FSL and FQL have acceptable performance in terms of adjusted number of virtual machine targeted to optimize SLA compliance and response time.

Pooyan Jamshidi, Miguel Velez, Christian Kästner et al.

Modern software systems are built to be used in dynamic environments using configuration capabilities to adapt to changes and external uncertainties. In a self-adaptation context, we are often interested in reasoning about the performance of the systems under different configurations. Usually, we learn a black-box model based on real measurements to predict the performance of the system given a specific configuration. However, as modern systems become more complex, there are many configuration parameters that may interact and we end up learning an exponentially large configuration space. Naturally, this does not scale when relying on real measurements in the actual changing environment. We propose a different solution: Instead of taking the measurements from the real system, we learn the model using samples from other sources, such as simulators that approximate performance of the real system at low cost. We define a cost model that transform the traditional view of model learning into a multi-objective problem that not only takes into account model accuracy but also measurements effort as well. We evaluate our cost-aware transfer learning solution using real-world configurable software including (i) a robotic system, (ii) 3 different stream processing applications, and (iii) a NoSQL database system. The experimental results demonstrate that our approach can achieve (a) a high prediction accuracy, as well as (b) a high model reliability.

Andreas Brunnert, Andre van Hoorn, Felix Willnecker et al.

DevOps is a trend towards a tighter integration between development (Dev) and operations (Ops) teams. The need for such an integration is driven by the requirement to continuously adapt enterprise applications (EAs) to changes in the business environment. As of today, DevOps concepts have been primarily introduced to ensure a constant flow of features and bug fixes into new releases from a functional perspective. In order to integrate a non-functional perspective into these DevOps concepts this report focuses on tools, activities, and processes to ensure one of the most important quality attributes of a software system, namely performance. Performance describes system properties concerning its timeliness and use of resources. Common metrics are response time, throughput, and resource utilization. Performance goals for EAs are typically defined by setting upper and/or lower bounds for these metrics and specific business transactions. In order to ensure that such performance goals can be met, several activities are required during development and operation of these systems as well as during the transition from Dev to Ops. Activities during development are typically summarized by the term Software Performance Engineering (SPE), whereas activities during operations are called Application Performance Management (APM). SPE and APM were historically tackled independently from each other, but the newly emerging DevOps concepts require and enable a tighter integration between both activity streams. This report presents existing solutions to support this integration as well as open research challenges in this area.

Armin Balalaie, Abbas Heydarnoori, Pooyan Jamshidi

Migration to the cloud has been a popular topic in industry and academia in recent years. Despite many benefits that the cloud presents, such as high availability and scalability, most of the on-premise application architectures are not ready to fully exploit the benefits of this environment, and adapting them to this environment is a non-trivial task. Microservices have appeared recently as novel architectural styles that are native to the cloud. These cloud-native architectures can facilitate migrating on-premise architectures to fully benefit from the cloud environments because non-functional attributes, like scalability, are inherent in this style. The existing approaches on cloud migration does not mostly consider cloud-native architectures as their first-class citizens. As a result, the final product may not meet its primary drivers for migration. In this paper, we intend to report our experience and lessons learned in an ongoing project on migrating a monolithic on-premise software architecture to microservices. We concluded that microservices is not a one-fit-all solution as it introduces new complexities to the system, and many factors, such as distribution complexities, should be considered before adopting this style. However, if adopted in a context that needs high flexibility in terms of scalability and availability, it can deliver its promised benefits.

Pooyan Jamshidi, Amir Sharifloo, Claus Pahl et al.

Cloud controllers aim at responding to application demands by automatically scaling the compute resources at runtime to meet performance guarantees and minimize resource costs. Existing cloud controllers often resort to scaling strategies that are codified as a set of adaptation rules. However, for a cloud provider, applications running on top of the cloud infrastructure are more or less black-boxes, making it difficult at design time to define optimal or pre-emptive adaptation rules. Thus, the burden of taking adaptation decisions often is delegated to the cloud application. Yet, in most cases, application developers in turn have limited knowledge of the cloud infrastructure. In this paper, we propose learning adaptation rules during runtime. To this end, we introduce FQL4KE, a self-learning fuzzy cloud controller. In particular, FQL4KE learns and modifies fuzzy rules at runtime. The benefit is that for designing cloud controllers, we do not have to rely solely on precise design-time knowledge, which may be difficult to acquire. FQL4KE empowers users to specify cloud controllers by simply adjusting weights representing priorities in system goals instead of specifying complex adaptation rules. The applicability of FQL4KE has been experimentally assessed as part of the cloud application framework ElasticBench. The experimental results indicate that FQL4KE outperforms our previously developed fuzzy controller without learning mechanisms and the native Azure auto-scaling.