A tensor intermediate representation for machine learning systems

Fatal accidents are a major issue hindering the wide acceptance of safety-critical systems that employ machine learning and deep learning models, such as automated driving vehicles. In order to use machine learning in a safety-critical system, it is necessary to demonstrate the safety and security of the system through engineering processes. However, thus far, no such widely accepted engineering concepts or frameworks have been established for these systems. The key to using a machine learning model in a deductively engineered system is decomposing the data-driven training of machine learning models into requirement, design, and verification, particularly for machine learning models used in safety-critical systems. Simultaneously, open problems and relevant technical fields are not organized in a manner that enables researchers to select a theme and work on it. In this study, we identify, classify, and explore the open problems in engineering (safety-critical) machine learning systems—that is, in terms of requirement, design, and verification of machine learning models and systems—as well as discuss related works and research directions, using automated driving vehicles as an example. Our results show that machine learning models are characterized by a lack of requirements specification, lack of design specification, lack of interpretability, and lack of robustness. We also perform a gap analysis on a conventional system quality standard SQuaRE with the characteristics of machine learning models to study quality models for machine learning systems. We find that a lack of requirements specification and lack of robustness have the greatest impact on conventional quality models.

Phenomics is a technology-driven approach with promising future to obtain unbiased data of biological systems. Image acquisition is relatively simple. However data handling and analysis are not as developed compared to the sampling capacities. We present a system based on machine learning (ML) algorithms and computer vision intended to solve the automatic phenotype data analysis in plant material. We developed a growth-chamber able to accommodate species of various sizes. Night image acquisition requires near infrared lightning. For the ML process, we tested three different algorithms: k-nearest neighbour (kNN), Naive Bayes Classifier (NBC), and Support Vector Machine. Each ML algorithm was executed with different kernel functions and they were trained with raw data and two types of data normalisation. Different metrics were computed to determine the optimal configuration of the machine learning algorithms. We obtained a performance of 99.31% in kNN for RGB images and a 99.34% in SVM for NIR. Our results show that ML techniques can speed up phenomic data analysis. Furthermore, both RGB and NIR images can be segmented successfully but may require different ML algorithms for segmentation.

Limitations of the chest X-ray (CXR) have resulted in attempts to create machine learning systems to assist clinicians and improve interpretation accuracy. An understanding of the capabilities and limitations of modern machine learning systems is necessary for clinicians as these tools begin to permeate practice. This systematic review aimed to provide an overview of machine learning applications designed to facilitate CXR interpretation. A systematic search strategy was executed to identify research into machine learning algorithms capable of detecting >2 radiographic findings on CXRs published between January 2020 and September 2022. Model details and study characteristics, including risk of bias and quality, were summarized. Initially, 2248 articles were retrieved, with 46 included in the final review. Published models demonstrated strong standalone performance and were typically as accurate, or more accurate, than radiologists or non-radiologist clinicians. Multiple studies demonstrated an improvement in the clinical finding classification performance of clinicians when models acted as a diagnostic assistance device. Device performance was compared with that of clinicians in 30% of studies, while effects on clinical perception and diagnosis were evaluated in 19%. Only one study was prospectively run. On average, 128,662 images were used to train and validate models. Most classified less than eight clinical findings, while the three most comprehensive models classified 54, 72, and 124 findings. This review suggests that machine learning devices designed to facilitate CXR interpretation perform strongly, improve the detection performance of clinicians, and improve the efficiency of radiology workflow. Several limitations were identified, and clinician involvement and expertise will be key to driving the safe implementation of quality CXR machine learning systems.

There is growing recognition that machine learning (ML) exposes new security and privacy vulnerabilities in software systems, yet the technical community's understanding of the nature and extent of these vulnerabilities remains limited but expanding. In this talk, we explore the threat model space of ML algorithms through the lens of Saltzer and Schroeder's principles for the design of secure computer systems. This characterization of the threat space prompts an investigation of current and future research directions. We structure our discussion around three of these directions, which we believe are likely to lead to significant progress. The first seeks to design mechanisms for assembling reliable records of compromise that would help understand the degree to which vulnerabilities are exploited by adversaries, as well as favor psychological acceptability of machine learning applications. The second encompasses a spectrum of approaches to input verification and mediation, which is a prerequisite to enable fail-safe defaults in machine learning systems. The third pursues formal frameworks for security and privacy in machine learning, which we argue should strive to align machine learning goals such as generalization with security and privacy desirata like robustness or privacy. Key insights resulting from these three directions pursued both in the ML and security communities are identified and the effectiveness of approaches are related to structural elements of ML algorithms and the data used to train them. We conclude by systematizing best practices in our growing community.

The research aims to identify a suitable machine learning model between various machine learning (ML) systems analysed. The input data are factors identified through principal component analysis (PCA) of potential entrepreneurs in the B40 category by analysing 1000 responses from this group through a survey instrument. The following machine learning systems are tested; Random Forest, Extra Trees, K-Neighbots, SVC, Ridge Classifier, Logistic Regression and Decision Tree to select and identify a case-based reasoning artificial intelligence (AI) system best suited in this scenario. Given the data set size, results based on accuracy indicate the best algorithm is Logistic Regression.

Evidence-based clinical engineering: Machine learning algorithms for prediction of defibrillator performance

Strategic Environmental Assessment (SEA) Process towards the Sustainable Design and Construction of Computer, Communication, Network Engineering, Machine Learning and Artificial Intelligence Systems

An analysis of agricultural resource management and the optimization of systems through the Intergration of advanced algorithms primarily revolves around the examination and the implementation of efficient innovative and computational techniques intended at enhancing agricultural productivity. Given this, the integration and use of algorithms that span from genetic algorithms, machine learning, and other artificial intelligence systems and methods that provide room for pest control, resource allocation, and the element of yield prediction. Advanced algorithms play a fundamental role in the facilitation of real-time data processing as well as decision making driven towards enabling farmers to focus on the optimization of fertilization, water usage, and crop rotation processes and schedules. Machine learning systems and models are further trained on real-time and historical data to predict pest infestations, weather patterns, and crop diseases-thus providing a proactive approach to the minimization of losses. On the other hand, genetic algorithms are further employed in the development of optimal planting methods while improving crop varieties, hence ensuring higher productivity and resilience. Further, these systems widely contribute to the process of sustainable farming procedures through the minimization of wastage and the promotion of eco-friendly agricultural practices. Through the analysis of weather conditions, soil health, and crop requirements, the designed algorithms ensure proper use of resources, resulting in increased profitability and cost saving practices for farmers. This article therefore seeks to conduct an analysis on agricultural resource management and optimization systems through the integration of advanced algorithms.

The aging and multimorbid population and health personnel shortages pose a substantial burden on primary health care. While predictive machine learning (ML) algorithms have the potential to address these challenges, concerns include transparency and insufficient reporting of model validation and effectiveness of the implementation in the clinical workflow. To systematically identify predictive ML algorithms implemented in primary care from peer-reviewed literature and US Food and Drug Administration (FDA) and Conformité Européene (CE) registration databases and to ascertain the public availability of evidence, including peer-reviewed literature, gray literature, and technical reports across the artificial intelligence (AI) life cycle. PubMed, Embase, Web of Science, Cochrane Library, Emcare, Academic Search Premier, IEEE Xplore, ACM Digital Library, MathSciNet, AAAI.org (Association for the Advancement of Artificial Intelligence), arXiv, Epistemonikos, PsycINFO, and Google Scholar were searched for studies published between January 2000 and July 2023, with search terms that were related to AI, primary care, and implementation. The search extended to CE-marked or FDA-approved predictive ML algorithms obtained from relevant registration databases. Three reviewers gathered subsequent evidence involving strategies such as product searches, exploration of references, manufacturer website visits, and direct inquiries to authors and product owners. The extent to which the evidence for each predictive ML algorithm aligned with the Dutch AI predictive algorithm (AIPA) guideline requirements was assessed per AI life cycle phase, producing evidence availability scores. The systematic search identified 43 predictive ML algorithms, of which 25 were commercially available and CE-marked or FDA-approved. The predictive ML algorithms spanned multiple clinical domains, but most (27 [63%]) focused on cardiovascular diseases and diabetes. Most (35 [81%]) were published within the past 5 years. The availability of evidence varied across different phases of the predictive ML algorithm life cycle, with evidence being reported the least for phase 1 (preparation) and phase 5 (impact assessment) (19% and 30%, respectively). Twelve (28%) predictive ML algorithms achieved approximately half of their maximum individual evidence availability score. Overall, predictive ML algorithms from peer-reviewed literature showed higher evidence availability compared with those from FDA-approved or CE-marked databases (45% vs 29%). The findings indicate an urgent need to improve the availability of evidence regarding the predictive ML algorithms' quality criteria. Adopting the Dutch AIPA guideline could facilitate transparent and consistent reporting of the quality criteria that could foster trust among end users and facilitating large-scale implementation.

Machine Learning (ML) systems such as Convolutional Neural Networks (CNNs) are susceptible to adversarial scenarios. In these scenarios, an attacker attempts to manipulate or deceive a machine learning model by providing malicious input resulting in incorrect predictions or decisions, which can have severe consequences in security, healthcare, and finance applications. Failures in the ML algorithm can lead to failures in the application domain and the system to which they provide functionality, which may possess performance requirements, hence the need for software reliability and resilience assessment. Many studies propose enhanced techniques to improve the robustness and resilience of ML algorithms. Yet, very few provide quantitative methods that promote risk assessment or measure progress toward improvements in such systems. This research demonstrates the applicability of software reliability and resilience tools to ML algorithms providing an objective approach to assess reliability and resilience recovery after a degradation from known adversarial attacks. An image classification model using CNNs has been created as the target of two generative adversarial attacks (i) the Fast Gradient Sign Method and (ii) the Projected Gradient Descent. The classifier and adversarial attacks were utilized in a min-max game scenario to replicate incremental learning and adaptive adversarial training, to minimize/maximize the number of failures of the classification algorithm. Model-generated data was collected and prepared in a format suitable for software reliability growth models (SRGM) with and without covariates and resilience models, including failure counts, memory usage, and the intensity of noise added by the attacks. Our results suggest that tools incorporating SRGMs and resilience techniques are suitable for quantifying and predicting the resilience and reliability of ML models and may find practical use in many domains in which these models are applied.

Analog/mixed-signal machine learning (ML) accelerators exploit the unique computing capability of analog/mixed-signal circuits and inherent error tolerance of ML algorithms to obtain higher energy efficiencies than digital ML accelerators. Unfortunately, these analog/mixed-signal ML accelerators lack programmability, and even instruction set interfaces, to support diverse ML algorithms or to enable essential software control over the energy-vs-accuracy tradeoffs. We propose PROMISE, the first end-to-end design of a PROgrammable MIxed-Signal accElerator from Instruction Set Architecture (ISA) to high-level language compiler for acceleration of diverse ML algorithms. We first identify prevalent operations in widely-used ML algorithms and key constraints in supporting these operations for a programmable mixed-signal accelerator. Second, based on that analysis, we propose an ISA with a PROMISE architecture built with silicon-validated components for mixed-signal operations. Third, we develop a compiler that can take a ML algorithm described in a high-level programming language (Julia) and generate PROMISE code, with an IR design that is both language-neutral and abstracts away unnecessary hardware details. Fourth, we show how the compiler can map an application-level error tolerance specification for neural network applications down to low-level hardware parameters (swing voltages for each application Task) to minimize energy consumption. Our experiments show that PROMISE can accelerate diverse ML algorithms with energy efficiency competitive even with fixed-function digital ASICs for specific ML algorithms, and the compiler optimization achieves significant additional energy savings even for only 1% extra errors.

In recent years, machine learning algorithms, or rather deep learning algorithms, have beenwidely used in many fields, including cybersecurity. However, machine learning systems are vulnerableto attacks by attackers, and this limits the use of machine learning, especially in nonstationaryenvironments with hostile actions, such as the cybersecurity field, where real attackersexist (for example, malware developers). With the rapid development of artificial intelligence (AI)and deep learning (GO) methods, it is important to ensure the safety and reliability of the implementedalgorithms. Recently, the vulnerability of deep learning algorithms to conflicting patternshas been widely recognized. Fabricated samples for analysis can lead to various violations of thebehavior of deep learning models, while people will consider them safe to use. The successful implementationof enemy attacks in real physical situations and scenarios of the real physical worldonce again proves their practicality. As a result, methods of adversarial attack and defense areattracting increasing attention from the security and machine learning communities and havebecome a hot topic of research in recent years not only in Russia, but also in other countries.Sberbank, Yandex, T1 Group, Atlas Medical Center and many others are developing competitivesolutions, including on the international market. Unfortunately, in the list of the 10 largest ITcompanies, the direction of Big Data, in particular, and protection against attacks is representedonly by the T1 Group company, but the market growth potential is huge. In this paper, the theoreticalfoundations, algorithms and application of methods of adversarial attacks of the enemy arepresented. Then a number of research papers on protection methods are described, covering awide range of research in this area. This article explores and summarizes adversarial attacks anddefenses, which represent the most up-to-date research in this field and meet the latest requirementsfor information security.

The purpose of this analysis is to assess the use of machine learning (ML) algorithms in the prediction of postoperative outcomes, including complications, recurrence, and death in transsphenoidal surgery. Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, we systematically reviewed all papers that used at least one ML algorithm to predict outcomes after transsphenoidal surgery. We searched Scopus, PubMed, and Web of Science databases for studies published prior to May 12, 2021. We identified 13 studies enrolling 5,048 patients. We extracted the general characteristics of each study; the sensitivity, specificity, area under the curve (AUC) of the ML models developed as well as the features identified as important by the ML models. We identified 12 studies with 5,048 patients that included ML algorithms for adenomas, three with 1807 patients specifically for acromegaly, and five with 2105 patients specifically for Cushing's disease. Nearly all were single-institution studies. The studies used a heterogeneous mix of ML algorithms and features to build predictive models. All papers reported an AUC greater than 0.7, which indicates clinical utility. ML algorithms have the potential to predict postoperative outcomes of transsphenoidal surgery and can improve patient care. Ensemble algorithms and neural networks were often top performers when compared with other ML algorithms. Biochemical and preoperative features were most likely to be selected as important by ML models. Inexplicability remains a challenge, but algorithms such as local interpretable model–agnostic explanation or Shapley value can increase explainability of ML algorithms. Our analysis shows that ML algorithms have the potential to greatly assist surgeons in clinical decision making.

With the wide application of machine learning technology, more and more sensitive data were used to develop the machine learning model. Unfortunately, machine learning systems have been shown to be vulnerable to various privacy leakage attacks, such as membership inference attacks and model inversion attacks. These privacy threats break down the confidentiality of the training data; what’s more, the leakage of sensitive privacy will reduce the enthusiasm of data owners for sharing data with the machine learning system. Without enough training data, more seriously, it will hinder the application of machine learning. Therefore, it requires analyzing strategies for privacy risks to help data owners pre-assess the candidate dataset and assist data owners in implementing reasonable privacy control. However, the systematic privacy risk assessment is still absent from the data owner’s perspective.This paper investigates and analyzes machine learning privacy risks to understand the relationship between training data properties and privacy leakage. Based on this recognition, we introduce a privacy risk assessment scheme based on the clustering distance of training data. Our clustering distance-based method can reflect the privacy risk level of a different individual data record. And then, we combine both existing privacy analysis based on data features and our clustering distance-based method to investigate privacy risks systematically. Our experiments showed that our clustering distance method and other set properties are tightly related to privacy leakage. And data owners can pre-assess the privacy risks of candidate datasets before uploading or sharing them to the machine learning model. If needed, decide to re-choose the dataset to reduce privacy risks to an acceptable level.

Although there is a large corpus of research focused on using machine learning to detect cyber threats, the solutions presented are rarely actually adopted in the real world. In this paper, we discuss the challenges that currently limit the adoption of machine learning in security operations, with a special focus on label acquisition, model deployment, and the integration of model findings into existing investigation workflows. Moreover, we posit that the conventional approach to the development of machine learning models, whereby researchers work offline on representative datasets to develop accurate models, is not valid for many cybersecurity use cases. Instead, a different approach is needed: to integrate the creation and maintenance of machine learning models into security operations themselves.