Autonomous Machine Learning Research Articles

Autonomous Machine Learning Algorithm for Stress Monitoring in Concrete Using Elastoacoustical Effect.

The measurement of stress in concrete structures is a complex issue. This paper presents a new measurement system called a self-acoustic system (SAS), which uses frequency measurements of acoustic waves to determine the condition of concrete structures. The SAS uses a positive feedback loop between ultrasonic heads, which causes excitation to a stable limit cycle. The frequency of this cycle is related to the propagation time of an acoustic wave, which directly depends on stresses in the test object. The coupling mechanism between acoustic wave propagation speed and stress is the elastoacoustic effect described in this paper. Thus, the proposed system enables the coupling between the limit cycle frequency and the stress degree of the concrete structure. This paper presents a machine learning algorithm to analyse the frequency spectrum of the SAS system. The proposed solution is a real-time classifier that enables online analysis of the frequency spectrum from the SAS system. With this approach, an autonomous system for stress condition identification of concrete structures is built and described.

Applying the FFP Approach to Wider Land Management Functions

The initial focus of implementing the Fit-for-Purpose Land Administration (FFPLA) methodology was to address the significant, global security of tenure divide. We argue that this land tenure methodology is proving successful in scaling up the provision of security of tenure for developing countries. The increasing adoption of the FFPLA methodology has also opened opportunities and provided flexibility for the innovative use of emerging technologies to accelerate the global roll out of security of tenure, such as the use of autonomous drones and machine learning techniques applied to image analysis. Despite wider adoption of participatory approaches to the recording of land tenure, similar FFP solutions for the other components of land administration services (land value, land use and land development) and land management functions are still evolving. This article therefore explores how the FFP approach can be applied to this wider set of land administration services and land management functions. A case study methodology, using three case studies, is used to determine if the case study approaches meet the FFP criteria. The focus is on the urban environment, drawing mostly from experiences and case studies in the Urban, Disaster Risk Management, Resilience & Land Global Practice of the World Bank. These opportunities for the wider application of the FFP approach and associated principles are being triggered by the innovative use of emerging new data capture technology developments. The paper examines the innovative use of these emerging technologies to identify a common set of data capture techniques and geospatial data that can be shared across a range of urban land administration and management activities. Finally, the paper discusses how individual land projects could be integrated into a more holistic land administration and management program approach and deliver a significant set of socio-economic benefits more quickly. It is found that the FFP approach can be more widely adopted across land administration and land management and in many cases can share a common set of geospatial data. The authors argue that the wider adoption and integration of these new, innovative FFP urban management approaches will require a significant cultural, professional, and institutional change from all stakeholders. Future work will explore more deeply these institutional weaknesses, which will provide a basis for guidance to the World Bank and similar institutions.

Recovery of meteorites using an autonomous drone and machine learning

AbstractThe recovery of freshly fallen meteorites from tracked and triangulated meteors is critical to determining their source asteroid families. Even though our ability to locate meteorite falls continues to improve, the recovery of meteorites remains a challenge due to large search areas with terrain and vegetation obscuration. To improve the efficiency of meteorite recovery, we have tested the hypothesis that meteorites can be located using machine learning techniques and an autonomous drone. To locate meteorites autonomously, a quadcopter drone first conducts a grid survey acquiring top‐down images of the strewn field from a low altitude. The drone‐acquired images are then analyzed using a machine learning classifier to identify meteorite candidates for follow‐up examination. Here, we describe a proof‐of‐concept meteorite classifier that deploys off‐line a combination of different convolution neural networks to recognize meteorites from images taken by drones in the field. The system was implemented in a conceptual drone setup and tested in the suspected strewn field of a recent meteorite fall near Walker Lake, Nevada.

Learning to recognize while learning to speak: Self-supervision and developing a speaking motor

Traditionally, learning speech synthesis and speech recognition were investigated as two separate tasks. This separation hinders incremental development for concurrent synthesis and recognition, where partially-learned synthesis and partially-learned recognition must help each other throughout lifelong learning. This work is a paradigm shift—we treat synthesis and recognition as two intertwined aspects of a lifelong learning agent. Furthermore, in contrast to existing recognition or synthesis systems, babies do not need their mothers to directly supervise their vocal tracts at every moment during the learning. We argue that self-generated non-symbolic states/actions at fine-grained time level help such a learner as necessary temporal contexts. Here, we approach a new and challenging problem—how to enable an autonomous learning system to develop an artificial speaking motor for generating temporally-dense (e.g., frame-wise) actions on the fly without human handcrafting a set of symbolic states. The self-generated states/actions are Muscles-like, High-dimensional, Temporally-dense and Globally-smooth (MHTG), so that these states/actions are directly attended for concurrent synthesis and recognition for each time frame. Human teachers are relieved from supervising learner’s motor ends. The Candid Covariance-free Incremental (CCI) Principal Component Analysis (PCA) is applied to develop such an artificial speaking motor where PCA features drive the motor. Since each life must develop normally, each Developmental Network-2 (DN-2) reaches the same network (maximum likelihood, ML) regardless of randomly initialized weights, where ML is not just for a function approximator but rather an emergent Turing Machine. The machine-synthesized sounds are evaluated by both the neural network and humans with recognition experiments. Our experimental results showed learning-to-synthesize and learning-to-recognize-through-synthesis for phonemes. This work corresponds to a key step toward our goal to close a great gap toward fully autonomous machine learning directly from the physical world.

Optimized and autonomous machine learning framework for characterizing pores, particles, grains and grain boundaries in microstructural images

Additively manufactured metals exhibit heterogeneous microstructure which dictates their material and failure properties. Experimental microstructural characterization techniques generate a large amount of data that requires expensive computational resources. In this work, an optimized machine learning (ML) framework is proposed to autonomously and efficiently characterize pores, particles, grains and grain boundaries (GBs) from a given microstructure image. First, using a classifier Convolutional Neural Network (CNN), defects such as pores, powder particles, or GBs were recognized from a given microstructure. Depending on the type of defect, two different processes were used. For powder particles or pores, binary segmentations were generated using an optimized Convolutional Encoder-Decoder Network (CEDN). The binary segmentations were used to obtain particle and pore size and location using an object detection ML network (YOLOv5). For GBs, another optimized CEDN was developed to generate RGB segmentation images, which were used to obtain grain size distribution using two regression CNNs. To optimize the RGB CEDN, the Deep Emulator Network SEarch (DENSE) method which employs the Covariance Matrix Adaptation - Evolution Strategy (CMA-ES) was implemented. The optimized RGB segmentation network showed a substantial reduction in training time and GPU usage compared to the unoptimized network, while maintaining high accuracy. Lastly, the proposed framework showed a significant improvement in analysis time when compared to conventional methods.

Crop and weed classification based on AutoML

<abstract> <p>CNN models already play an important role in classification of crop and weed with high accuracy, more than 95% as reported in literature. However, to manually choose and fine-tune the deep learning models becomes laborious and indispensable in most traditional practices and research. Moreover, the classic objective functions are not thoroughly compatible with agricultural farming tasks as the corresponding models suffer from misclassifying crop to weed, often more likely than in other deep learning application domains. In this paper, we applied autonomous machine learning with a new objective function for crop and weed classification, achieving higher accuracy and lower crop killing rate (rate of identifying a crop as a weed). The experimental results show that our method outperforms state-of-the-art applications, for example, ResNet and VGG19.</p> </abstract>

Employing and Interpreting a Machine Learning Target-Cognizant Technique for Analysis of Unknown Signals in Multiple Reaction Monitoring.

The aim of this interdisciplinary work is a robust signal processing and autonomous machine learning framework to associate well-known (target) as well as any potentially unknown (non-target) peaks present within gas chromatography-mass spectrometry (GC/MS/MS) raw instrument signal. Particularly, this work evaluates three machine learning algorithms abilities to autonomously associate raw signal peaks based on accuracy in training and testing. A target is a known congener that is expected to be present within the raw instrument signal and a non-target is an unknown or unexpected compound. Autonomously identifying target peaks within the GC/MS/MS and associating them with non-target peaks can help improve the analysis of collected samples. Association of peaks refers to classifying peaks as known congeners regardless if the peak is a target or non-target. Uncertainty of peaks fitted and discovered through raw instrument signals from GC/MS/MS data is assessed to create topographical illustrations of target annotated peaks among sample raw instrument signals collected across diverse locations in the Chicago area. The term “annotated peak” is used to assign peaks found at specific retention times as a known congener. Adaptive signal processing techniques are utilized to smooth data and correct baseline drifts as well as detect and separate coeluted (overlapped) peaks in the raw instrument signal to provide key feature extraction. 150 air samples are analyzed for individual polychlorinated biphenyls (PCB) with GC/MS/MS across Chicago, IL. 80% of the data is used for training classification of target PCBs and 20% of the data is evaluated to identify and associate consistently occurring non-target peaks with target PCBs. A random forest classifier is used to associate identified peaks to target PCB peaks. Geographical topographical representations of target PCBs in the raw instrument signal demonstrates how PCBs accumulate and degrade in certain locations.

Hip Fracture Risk Assessment in Elderly and Diabetic Patients: Combining Autonomous Finite Element Analysis and Machine Learning.

Autonomous finite element analyses (AFE) based on CT scans predict the biomechanical response of femurs during stance and sidewise fall positions. We combine AFE with patient data via a machine learning (ML) algorithm to predict the risk of hip fracture. An opportunistic retrospective clinical study of CT scans is presented, aimed at developing a ML algorithm with AFE for hip fracture risk assessment in type 2 diabetic mellitus (T2DM) and non-T2DM patients. Abdominal/pelvis CT scans of patients who experienced a hip fracture within 2 years after an index CT scan were retrieved from a tertiary medical center database. A control group of patients without a known hip fracture for at least 5 years after an index CT scan was retrieved. Scans belonging to patients with/without T2DM were identified from coded diagnoses. All femurs underwent an AFE under three physiological loads. AFE results, patient's age, weight, and height were input to the ML algorithm (support vector machine [SVM]), trained by 80% of the known fracture outcomes, with cross-validation, and verified by the other 20%. In total, 45% of available abdominal/pelvic CT scans were appropriate for AFE (at least 1/4 of the proximal femur was visible in the scan). The AFE success rate in automatically analyzing CT scans was 91%: 836 femurs we successfully analyzed, and the results were processed by the SVM algorithm. A total of 282 T2DM femurs (118 intact and 164 fractured) and 554 non-T2DM (314 intact and 240 fractured) were identified. Among T2DM patients, the outcome was: Sensitivity 92%, Specificity 88% (cross-validation area under the curve [AUC] 0.92) and for the non-T2DM patients: Sensitivity 83%, Specificity 84% (cross-validation AUC 0.84). Combining AFE data with a ML algorithm provides an unprecedented prediction accuracy for the risk of hip fracture in T2DM and non-T2DM populations. The fully autonomous algorithm can be applied as an opportunistic process for hip fracture risk assessment. © 2023 The Authors. Journal of Bone and Mineral Research published by Wiley Periodicals LLC on behalf of American Society for Bone and Mineral Research (ASBMR).

From the Tree of Knowledge & the Golem of Prague: The Ethics of Artificial Intelligence Through Jewish Eyes

This paper discusses the regulation of artificial intelligence from a Jewish perspective, with an emphasis on the regulation of machine learning and its application to autonomous vehicles and machine learning. Through the Biblical story of Adam and Eve as well as Golem legends from Jewish folklore, we derive several basic principles that underlie a Jewish perspective on the moral and legal person-hood of robots and other artificially intelligent agents. We argue that religious ethics in general, and Jewish ethics in particular, show us that the dangers of granting moral person-hood to robots and in particular to autonomous vehicles lie not in the fact that they lack a soul – or consciousness or feelings or interests – but because to do so weakens our own ability to develop as fully autonomous legal and moral persons. Instead, we argue that existing legal persons should continue to maintain legal control over artificial agents, while natural persons assume ultimate moral responsibility for choices made by artificial agents they employ in their service. In the final section of the paper we discuss the trolley dilemma in the context of governing autonomous vehicles and sketch out an application of Jewish ethics in a case where we are asking Artificial Intelligence to make life and death decisions. Our novel contribution is twofold; first, we bring a religious approach to the discussion of the ethics of Artificial Intelligence which has hitherto been dominated by secular Western philosophies; second, we raise the idea that artificial entities who are trained through machine learning can be ethically trained in much the same way that human are – through reading and reflecting on core religious texts. This is both a way of ensuring the ethical regulation of artificial intelligence, but also promotes other core values of regulation, such as democratic engagement and user choice.

What lies beneath: Material classification for autonomous excavators using proprioceptive force sensing and machine learning

The ability of robotic excavators to acquire meaningful knowledge about materials during digging can augment their autonomous functionality, as well as optimize downstream operations in construction and mining. Some material properties, such as rock sizes, can be determined visually, but these methods cannot see what lies beneath. In this work, a classification methodology that utilizes only proprioceptive force data acquired from an autonomous digging system and machine learning algorithms is proposed for excavation material identification. The consistent performance synonymous with autonomous digging systems allows for the use of basic features extracted from the force data for classification. A proof of concept of this novel approach to excavation material classification is demonstrated through a binary classification of rock and gravel materials. Force data were obtained from full-scale autonomous loading trials with a 14-tonne capacity load-haul-dump machine at a mining and construction test facility. Preliminary results achieved a classification accuracy of 90%.

Decentralized Privacy Using Blockchain-Enabled Federated Learning in Fog Computing

As the extension of cloud computing and a foundation of IoT, fog computing is experiencing fast prosperity because of its potential to mitigate some troublesome issues, such as network congestion, latency, and local autonomy. However, privacy issues and the subsequent inefficiency are dragging down the performances of fog computing. The majority of existing works hardly consider a reasonable balance between them while suffering from poisoning attacks. To address the aforementioned issues, we propose a novel blockchain-enabled federated learning (FL-Block) scheme to close the gap. FL-Block allows local learning updates of end devices exchanges with a blockchain-based global learning model, which is verified by miners. Built upon this, FL-Block enables the autonomous machine learning without any centralized authority to maintain the global model and coordinates by using a Proof-of-Work consensus mechanism of the blockchain. Furthermore, we analyze the latency performance of FL-Block and further derive the optimal block generation rate by taking communication, consensus delays, and computation cost into consideration. Extensive evaluation results show the superior performances of FL-Block from the aspects of privacy protection, efficiency, and resistance to the poisoning attack.

Applying seven resilience principles on the Vision of the Digital City

• We use an ecological lens on physiognomies of the digital city-development. • Using resilience principles can reduce risks in digitalization of city functions. • Strategically design for both buffering instant shocks and slow erosion • Use autonomic self-organizing machine learning for whole digital network resilience • Keep a separated and redundant analogue technology for critical infrastructures

뉴로모픽 모델과 기계학습 모델을 지원하는 자율형 IoT 프로그래밍 패러다임

센서 기술과 고속 통신 인프라의 발전으로 IoT 응용 서비스에 대한 요구와 기대가 증가하고 있다. 많은 센서가 운용되고 네트워크로 연결되더라도 센서 데이터를 모두 서버로 보내서 처리하는 것은 통신 대역 및 저장 공간 측면에서 비효율적이다. 한편, 인공지능 기술 발전에 따라 IoT의 지능적 처리에 대한 요구가 증가되고 있다. 본 논문에서는 IoT 서버 뿐만 아니라 클라이언트에서 뉴로모픽 기반 모델과 기계학습 모델 등을 사용할 수 있는 프로그래밍 패러다임에 대해서 제안한다. 제안한 프로그래밍 패러다임은 다양한 인공지능 모듈을 IoT 클라이언트와 서버 프로그램에서도 활용할 수 있는 패러다임이라는 점에서 향후 지능형 IoT뿐만 아니라 자율형 IoT 환경에서 유용할 것으로 기대된다.

From the Tree of Knowledge and the Golem of Prague to Kosher Autonomous Cars: The Ethics of Artificial Intelligence Through Jewish Eyes

Abstract This article discusses the regulation of artificial intelligence from a Jewish perspective, with an emphasis on the regulation of machine learning and its application to autonomous vehicles and machine learning. Through the Biblical story of Adam and Eve as well as Golem legends from Jewish folklore, we derive several basic principles that underlie a Jewish perspective on the moral and legal personhood of robots and other artificially intelligent agents. We argue that religious ethics in general, and Jewish ethics in particular, show us that the dangers of granting moral personhood to robots and in particular to autonomous vehicles lie not in the fact that they lack a soul—or consciousness or feelings or interests—but because to do so weakens our own ability to develop as fully autonomous legal and moral persons. Instead, we argue that existing legal persons should continue to maintain legal control over artificial agents, while natural persons assume ultimate moral responsibility for choices made by artificial agents they employ in their service. In the final section of the article we discuss the trolley dilemma in the context of governing autonomous vehicles and sketch out an application of Jewish ethics in a case where we are asking Artificial Intelligence to make life and death decisions. Our novel contribution is two-fold; first, we bring a religious approach to the discussion of the ethics of Artificial Intelligence which has hitherto been dominated by secular Western philosophies; second, we raise the idea that artificial entities who are trained through machine learning can be ethically trained in much the same way that human are—through reading and reflecting on core religious texts. This is both a way of ensuring the ethical regulation of artificial intelligence, but also promotes other core values of regulation, such as democratic engagement and user choice.

Prediction of Drag Force on Vehicles in a Platoon Configuration Using Machine Learning

Machine learning is used for extraction of valuable information from data thus helping in exploration of hidden patterns, leading to learning models that can be used for prediction. In the domain of autonomous vehicles machine learning techniques have been applied in several areas, vehicle platooning being one of them. Vehicle platooning is a vital feature of automated highways which provides the key benefits of fuel economy, road safety and environmental protection coupled with safe road transportation. However, high computational cost associated with the numerical simulation of vehicle aerodynamics makes the Computational Fluid Dynamics (CFD) study of vehicle platoon prohibitively expensive and complex. Machine learning, with its high predictive power, has emerged as a promising compliment to CFD studies of external aerodynamics. This paper presents estimation error based performance comparison of five different supervised learning algorithms: Support Vector Regression, Polynomial Regression, Linear Regression and two different models of Neural Networks for prediction of aerodynamic drag coefficient corresponding to each vehicle in a two, three and four vehicle platoon configurations based on the drag coefficients provided by experimental study at different inter-vehicle distances. Predicted drag coefficients are then juxtaposed with CFD data from numerical simulations to evaluate closeness to experimental drag coefficients. Results reveal that polynomial regression model best fits the aerodynamics with 0.0223 estimation error. To the best of our knowledge no machine learning based methods have been applied before for modeling aerodynamic drag on vehicle platoon.

Compelling Unification of Forward-Error Correction and the Turing Machine

Numerous specialists would concur that, had it not been for Boolean rationale, the perception of RAID may never have happened. In this work, we approve the change of rasterization, which exemplifies the broad standards of autonomous machine learning. In this position paper we depict a cacheable instrument for envisioning Boolean rationale (Jog), demonstrating that hash tables can be made Bayesian, marked, and helpful.

In Search of the Unseen: Towards Superhuman Intuition

AbstractAlisa Andrasek is a professor at RMIT University in Melbourne and previously director of Wonderlab at the Bartlett School of Architecture, University College London. Her work is a restless search for the beauty of novel architectural forms and techniques utilising autonomous agents, architectonic algorithms, machine learning and robotics.

The Necessity, Promise and Challenge of Automated Biodiversity Surveys

SummaryWe are in the midst of a transformation in the way that biodiversity is observed on the planet. The approach of direct human observation, combining efforts of both professional and citizen scientists, has recently generated unprecedented amounts of data on species distributions and populations. Within just a few years, however, we believe that these data will be swamped by indirect biodiversity observations that are generated by autonomous sensors and machine learning classification models. In this commentary, we discuss three important elements of this shift towards indirect, technology driven observations. First, we note that the biodiversity data sets available today cover a very small fraction of all places and times that could potentially be observed, which suggests the necessity of developing new approaches that can gather such data at even larger scales, with lower costs. Second, we highlight existing tools and efforts that are already available today to demonstrate the promise of automated methods to radically increase biodiversity data collection. Finally, we discuss one specific outstanding challenge in automated biodiversity survey methods, which is how to extract useful knowledge from observations that are uncertain in nature. Throughout, we focus on one particular type of biodiversity data - point occurrence records - that are frequently produced by citizen science projects, museum records and systematic biodiversity surveys. As indirect observation methods increase the spatiotemporal scope of these point occurrence records, ecologists and conservation biologists will be better able to predict shifting species distributions, track changes to populations over time and understand the drivers of biodiversity occurrence.

Autonomic machine learning platform

Acquiring information properly through machine learning requires familiarity with the available algorithms and understanding how they work and how to address the given problem in the best possible way. However, even for machine-learning experts in specific industrial fields, in order to predict and acquire information properly in different industrial fields, it is necessary to attempt several instances of trial and error to succeed with the application of machine learning. For non-experts, it is much more difficult to make accurate predictions through machine learning.In this paper, we propose an autonomic machine learning platform which provides the decision factors to be made during the developing of machine learning applications. In the proposed autonomic machine learning platform, machine learning processes are automated based on the specification of autonomic levels. This autonomic machine learning platform can be used to derive a high-quality learning result by minimizing experts’ interventions and reducing the number of design selections that require expert knowledge and intuition. We also demonstrate that the proposed autonomic machine learning platform is suitable for smart cities which typically require considerable amounts of security sensitive information.

Particle swarm optimization for generating interpretable fuzzy reinforcement learning policies

Fuzzy controllers are efficient and interpretable system controllers for continuous state and action spaces. To date, such controllers have been constructed manually or trained automatically either using expert-generated problem-specific cost functions or incorporating detailed knowledge about the optimal control strategy. Both requirements for automatic training processes are not found in most real-world reinforcement learning (RL) problems. In such applications, online learning is often prohibited for safety reasons because it requires exploration of the problem’s dynamics during policy training. We introduce a fuzzy particle swarm reinforcement learning (FPSRL) approach that can construct fuzzy RL policies solely by training parameters on world models that simulate real system dynamics. These world models are created by employing an autonomous machine learning technique that uses previously generated transition samples of a real system. To the best of our knowledge, this approach is the first to relate self-organizing fuzzy controllers to model-based batch RL. FPSRL is intended to solve problems in domains where online learning is prohibited, system dynamics are relatively easy to model from previously generated default policy transition samples, and it is expected that a relatively easily interpretable control policy exists. The efficiency of the proposed approach with problems from such domains is demonstrated using three standard RL benchmarks, i.e., mountain car, cart-pole balancing, and cart-pole swing-up. Our experimental results demonstrate high-performing, interpretable fuzzy policies.