Learning Iteration Research Articles

Designed Organizational Search: A Comparative Analysis of Alternative Procedures for Learning from Success

Sampling on the dependent variable is unlikely to be an effective way to learn and develop the strategy. Even so, organizations spend millions of dollars on processes such as Appreciative Inquiry that make inferences about how to adapt their strategies, routines, and practices based upon only successful examples. Two techniques that are common to this kind of learning process are searching solely for successful solutions and reframing search problems (e.g., unconditionally positive questions). We build a computational model by formalizing appreciative inquiry and comparing it with other, similar processes to understand their relative effectiveness. We find that the organizations simulated in our computational model almost always improved performance over time, despite learning solely from successful observations. Their relative effectiveness depended on the complexity of the problems, the number of iterations of learning, and how much the learning process preserved variety in potential solutions. These findings suggest that appreciative inquiry may be most effective when people take the cost and complexity of organizational problems into account before engaging in the learning process and adapt the process accordingly. These findings contribute to research on organizational learning by explaining why learners may benefit from structuring the way they communicate as they search, why reframing performance measures may dissolve search problems, and how designed organizational search enables managers to be more deliberate about organizational learning.

Safety Risk Evaluation of Tourism Management System Based on PSO-BP Neural Network

With the recovery of the tourism industry and the development of the field of artificial intelligence, the application of intelligent neural network technology to management systems for safety risk assessment is the choice of the times and an urgent real need for relevant practitioners. In this study, the BP neural network algorithm is used as a tool for safety evaluation of the tourism management system. The three-layer structure of the BP neural network and the role of nodes in it are introduced, the weight values and thresholds of nodes in each of the three layers are calculated, and the particle swarm algorithm is added to optimize the model. In the practical stage, the tourism data of a place in Switzerland was selected as the training data. After 12,859 iterations, the model achieved the best calibration error of 0.126. After 300 iterations of learning, the BP algorithm optimized with the PSO algorithm has a faster convergence rate, which indicates that the performance of the optimized algorithm has improved significantly and has the global search capability that was not available before, which significantly outperforms the FastText and LSTM models. With the increase in the number of samples of tourism data, macroaccuracy always remains above 80%, so it proves that the optimization algorithm used in this study is an effective and reliable model.

APPLICATION OF THE BACKPROPAGATION METHOD TO PREDICT RAINFALL IN NORTH SUMATRA PROVINCE

Natural disasters are to blame for the high level of community loss. This is due to the community's lack of information about potential disasters around them. As a result, public understanding of disaster response is extremely low. As a result, weather information is critical for the smooth operation of human activities and activities, such as determining the amount of rainfall. The goal of this research is to identify the best model for predicting rainfall in North Sumatra Province and to forecast rainfall trends for the coming year. The rainfall time series data used in this study were collected from six stations in North Sumatra Province over the last ten years, including the Sibolga Meteorological Station, Aek Godang Meteorological Station, and Silangit Meteorological Station. Backpropagation is used in this study. Backpropagation is one of the methods used in artificial neural networks, which are usually divided into three layers: an input layer, a hidden layer, and an output layer connected by weights. During the training stage, the learning rate, iteration, and number of nodes in the hidden layer were all tested. Following the training process, the best model will be used for testing. The best model was obtained using rainfall data from North Sumatra Province, with an optimal iteration of 1000 iterations, an optimal learning rate of 0.1 in the learning rate trial, and the best number of hidden 5 nodes. During the testing, the MSE values were 0.047 and 0.022, respectively, and the MSE squared value was 0.0022 and 0.00049.

A Hierarchical Reinforcement Learning Framework for UAV Path Planning in Tactical Environments

Tactical UAV path planning under radar threat using reinforcement learning involves particular challenges ranging from modeling related difficulties to sparse feedback problem. Learning goal-directed behavior with sparse feedback from complex environments is a fundamental challenge for reinforcement learning algorithms. In this paper we extend our previous work in this area to provide a solution to the problem setting stated above, using Hierarchical Reinforcement Learning (HRL) in a novel way that involves a meta controller for higher level goal assignment and a controller that determines the lower-level actions of the agent. Our meta controller is based on a regression model trained using a state transition scheme that defines the evolution of goal designation, whereas our lower-level controller is based on a Deep Q Network (DQN) and is trained via reinforcement learning iterations. This two-layer framework ensures that an optimal plan for a complex path, organized as multiple goals, is achieved gradually, through piecewise assignment of sub-goals, and thus as a result of a staged, efficient and rigorous procedure.

Anti-Jamming Strategy for Federated Learning in Internet of Medical Things: A Game Approach.

Federated learning (FL) is a new dawn of artificial intelligence (AI), in which machine learning models are constructed in a distributed manner while communicating only model parameters between a centralized aggregator and client internet-of-medical-things (IoMT) nodes. The performance of such a learning technique can be seriously hampered by the activities of a malicious jammer robot. In this paper, we study client selection and channel allocation along with the power control problem of the uplink FL process in IoMT domain under the presence of a jammer from the perspective of long-term learning duration. We map the interaction between the FL network and the jammer in each learning iteration as a Stackelberg game, in which the jammer acts as the leader and the FL network serves as the follower. We consider the client and channel selection as well as the power control jointly as the strategy of this game. Upon formulating the game, we find the joint best response strategy for both types of players by leveraging the difference of convex (DC) programming approach and the dual decomposition technique. Beside the availability of the complete information to both the players, we also study the problem from the perspective that the FL network knows the partial information of the other player. Extensive simulations have been conducted to verify the effectiveness of the proposed algorithms in the jamming game.

Identifying environmentally sustainable pavement management strategies via deep reinforcement learning

Pavement life cycle assessments (LCAs) enable decision-makers to evaluate the environmental impact of alternative maintenance, rehabilitation, and reconstruction strategies. This paper explores the viability of deep reinforcement learning (DRL), a framework that enables agents to learn optimal actions within a given situation, to identify environmentally benign pavement management strategies. More specifically, this study utilizes proximal-policy optimization (PPO), a subtype of DRL algorithms, to identify a management strategy that minimizes the expected global warming impact of a pavement facility over its lifecycle. Through an urban Interstate case study, this paper shows that the proposed PPO algorithm identifies management strategies that are anticipated to reduce the expected global warming impact of a pavement facility over its planning horizon by 16 percent relative to traditional practice. Furthermore, the PPO algorithm is able to identify this management strategy in only 25 learning iterations, which is in stark comparison to Q-learning, a common reinforcement learning algorithm, that requires 70,000 learning iterations. The results of this work highlight the viability of DRL to integrate within complex LCA models to determine environmentally sustainable pavement management strategies.

A Semi-Parallel Active Learning Method Based on Kriging for Structural Reliability Analysis

The reliability analysis system is currently evolving, and reliability analysis efforts are also focusing more on correctness and efficiency. The effectiveness of the active learning Kriging metamodel for the investigation of structural system reliability has been demonstrated. In order to effectively predict failure probability, a semi-parallel active learning method based on Kriging (SPAK) is developed in this study. The process creates a novel learning function called UA, which takes the correlation between training points and samples into account. The UA function has been developed from the U function but is distinct from it. The UA function improves the original U function, which pays too much attention to the area near the threshold and the accuracy of the surrogate model is improved. The semi-parallel learning method is then put forth, and it works since UA and U functions are correlated. One or two training points will be added sparingly during the model learning iteration. It effectively lowers the required training points and iteration durations and increases the effectiveness of model building. Finally, three numerical examples and one engineering application are carried out to show the precision and effectiveness of the suggested method. In application, evaluation efficiency is increased by at least 14.5% and iteration efficiency increased by 35.7%. It can be found that the proposed algorithm is valuable for engineering applications.

Multi-Generator Discriminator Network Using Texture-Edge Information

In the proposed paper, a parallel structure type Generative Adversarial Network (GAN) using edge and texture information is proposed. In the existing GAN-based model, many learning iterations had to be given to obtaining an output that was somewhat close to the original data, and noise and distortion occurred in the output image even when learning was performed. To solve this problem, the proposed model consists of two generators and three discriminators to propose a network in the form of a parallel structure. In the network, each edge information and texture information were received as inputs, learning was performed, and each character was combined and outputted through the Combine Discriminator. Through this, edge information and distortion of the output image were improved even with fewer iterations than DCGAN, which is the existing GAN-based model. As a result of learning on the network of the proposed model, a clear image with improved contour and distortion of objects in the image was output from about 50,000 iterations.

SAFELearning: Secure Aggregation in Federated Learning With Backdoor Detectability

For model privacy, local model parameters in federated learning shall be obfuscated before sent to the remote aggregator. This technique is referred to as <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">secure aggregation</i> . However, secure aggregation makes model poisoning attacks such as backdooring more convenient given that existing anomaly detection methods mostly require access to plaintext local models. This paper proposes a new federated learning technique SAFE-Learning to support backdoor detection for secure aggregation. We achieve this through two new primitives - <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">oblivious random grouping (ORG)</i> and <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">partial parameter disclosure (PPD)</i> . ORG partitions participants into one-time random subgroups with group configurations oblivious to participants; PPD allows secure partial disclosure of aggregated subgroup models for anomaly detection without leaking individual model privacy. ORG is based on our construction of several new primitives including tree-based random subgroup generation, oblivious secure aggregation, and randomized Diffie-Hellman key exchange. ORG can thwart colluding attackers from knowing each other’s group membership assignment with non-negligible advantage than random guess. Backdoor attacks are detected based on statistical distributions of the subgroup aggregated parameters of the learning iterations. SAFELearning can significantly reduce backdoor model accuracy without jeopardizing the main task accuracy under common backdoor strategies. Extensive experiments show SAFELearning is robust against malicious and faulty participants, whilst being more efficient than the state-of-art secure aggregation protocol in terms of both communication and computation costs.

A Novel Predefined Time PD-Type ILC Paradigm for Nonlinear Systems

Intelligent robotics has drawn a great deal of attention due to its high precision, stability, and reliability, which are the basic key factors for industrial automation. This paper proposes an iterative learning control (ILC) technique with predefined-time convergence as a solution to an applied engineering problem, namely, that local time cannot be preset when a second-order nonlinear system undertakes control of the accurate tracking of local time under any initial iterative value. A time-varying sliding surface with an initial value of zero was designed, and it was theoretically proven that the trajectory tracking error in the sliding surface could converge to zero within a predefined time. The iterative control problem of trajectory tracking was thus changed to an iterative control problem of time-varying sliding-mode surface tracing with a starting value of zero. A PD-type closed-loop ILC with a time-varying sliding mode surface was designed such that the trajectory tracking error converged and stabilized on the sliding mode surface after a finite number of learning iterations. The control goal for the system’s output was the ability to track the desired trajectory accurately within a predefined time interval, and it was achieved by combining this with the predefined time convergence characteristics of the time-varying sliding mode surface. Numerical simulation of trajectory tracking control of a repetitive motion manipulator was used to verify the effectiveness of the proposed controller and its robustness in the face of external disturbances.

Semi-Supervised Random Forest Methodology for Fault Diagnosis in Air-Handling Units

Air-handling units have been widely used in indoor air conditioning and circulation in modern buildings. The data-driven FDD method has been widely used in the field of industrial roads, and has been widely welcomed because of its extensiveness and flexibility in practical applications. Under the condition of sufficient labeled data, previous studies have verified the utility and value of various supervised learning algorithms in FDD tasks. However, in practice, obtaining sufficient labeled data can be very challenging, expensive, and will consume a lot of time and manpower, making it difficult or even impractical to fully explore the potential of supervised learning algorithms. To solve this problem, this study proposes a semi-supervised FDD method based on random forest. This method adopts a self-training strategy for semi-supervised learning and has been verified in two practical applications: fault diagnosis and fault detection. Through a large number of data experiments, the influence of key learning parameters is statistically represented, including the availability of marked data, the number of iterations of maximum half-supervised learning, and the threshold of utilization of pseudo-label data. The results show that the proposed method can effectively utilize a large number of unlabeled data, improve the generalization performance of the model, and improve the diagnostic accuracy of different column categories by about 10%. The results are helpful for the development of advanced data-driven fault detection and diagnosis tools for intelligent building systems.

Iterative learning contouring control: Theory and application to biaxial systems

This paper is concerned with the iterative learning control (ILC) for the contouring control of multi-axis motion system. A novel ILC algorithm, which is called iterative learning contouring control (ILCC), is proposed and verified with biaxial motion systems. In motion control, conventional ILC requires that the number of control objectives is the same as the number of command inputs, which is called square property. This allows for the formulation of transfer function or matrix inversion. The unique feature of contouring control is its non-square property. For contouring control, conventional ILC circumvents the non-square property by adopting the configuration of cross-coupled control (CCC). In other words, the estimated contour error is decomposed into components of individual axis for compensation by ILC and thus the contour error is reduced indirectly. The proposed ILCC can deal with the non-square property and can directly reduce the contour error by adopting the equivalent contour error as the control objective of the ILC. The equivalent contour error represents an accurate contour error model and hence better contouring accuracy can be achieved. The ILCC algorithm is implemented on two platforms, an XY table and a commercial CNC machine tool. Experiments on the contouring control of a circular path with different speeds and a B-spline path are conducted. It is found that the contour error can be reduced about 88% after several learning iterations. The experimental results confirm the effectiveness of the proposed method and show the excellent contouring performance.

Robust Inverse Q-Learning for Continuous-Time Linear Systems in Adversarial Environments.

This article proposes robust inverse Q -learning algorithms for a learner to mimic an expert's states and control inputs in the imitation learning problem. These two agents have different adversarial disturbances. To do the imitation, the learner must reconstruct the unknown expert cost function. The learner only observes the expert's control inputs and uses inverse Q -learning algorithms to reconstruct the unknown expert cost function. The inverse Q -learning algorithms are robust in that they are independent of the system model and allow for the different cost function parameters and disturbances between two agents. We first propose an offline inverse Q -learning algorithm which consists of two iterative learning loops: 1) an inner Q -learning iteration loop and 2) an outer iteration loop based on inverse optimal control. Then, based on this offline algorithm, we further develop an online inverse Q -learning algorithm such that the learner mimics the expert behaviors online with the real-time observation of the expert control inputs. This online computational method has four functional approximators: a critic approximator, two actor approximators, and a state-reward neural network (NN). It simultaneously approximates the parameters of Q -function and the learner state reward online. Convergence and stability proofs are rigorously studied to guarantee the algorithm performance.

Convergence of Finite Memory Q Learning for POMDPs and Near Optimality of Learned Policies Under Filter Stability

In this paper, for partially observed Markov decision problems (POMDPs), we provide the convergence of a Q learning algorithm for control policies using a finite history of past observations and control actions, and consequentially, we establish near optimality of such limit Q functions under explicit filter stability conditions. We present explicit error bounds relating the approximation error to the length of the finite history window. We establish the convergence of such Q learning iterations under mild ergodicity assumptions on the state process during the exploration phase. We further show that the limit fixed point equation gives an optimal solution for an approximate belief Markov decision problem (MDP). We then provide bounds on the performance of the policy obtained using the limit Q values compared with the performance of the optimal policy for the POMDP, in which we also present explicit conditions using recent results on filter stability in controlled POMDPs. Whereas there exist many experimental results, (i) the rigorous asymptotic convergence (to an approximate MDP value function) for such finite memory Q learning algorithms and (ii) the near optimality with an explicit rate of convergence (in the memory size) under filter stability are results that are new to the literature to our knowledge. Funding: This research was supported in part by the Natural Sciences and Engineering Research Council (NSERC) of Canada.

Blockchain-based decentralized federated transfer learning methodology for collaborative machinery fault diagnosis

Due to the limitations of data quality and quantity of a single industrial user, the development of intelligent machinery fault diagnosis methods has been reaching a bottleneck in the perspectives of both academic research and engineering applications in the recent years. Collaborative fault diagnosis model development has been receiving increasing attention, where the distributed data at different users are explored simultaneously. However, data security and privacy are the major industrial concerns, which have not been well addressed in the literature. In this paper, a blockchain-based decentralized federated transfer learning method is proposed for collaborative machinery fault diagnosis. A tailored committee consensus scheme is designed for optimization of the model aggregation process, and a source data-free transfer learning method is further proposed. After global model initialization, the fault diagnosis model can be built through iterations of committee member selection, performance evaluation, transfer learning, model aggregation and blockchain updates. The experiments on two decentralized fault diagnosis datasets are implemented for validations, and higher than 90% testing accuracies can be generally achieved. The experimental results indicate the proposed method is effective in data privacy-preserving collaborative fault diagnosis of multiple users. It offers a promising tool for applications in the real industrial scenarios.

DEEP LEARNING AND SUCCESSIVE ITERATIONS FOR AN INVERSE PROBLEM OF THE VARIABLE DENSITY RECONSTRUCTION BY THE BOUNDARY CONTROL METHOD

For the first time, two approaches, the deep learning and successive iterations, are proposed for use and implemented for enhancing the images reconstructed by the boundary control method. The construction of successive iterations is based on the reduction of a nonlinear inverse problem for a time-domain wave equation to a linear integral equation of the first kind. The computational effectiveness of the numerical techniques is demonstrated in the numerical experiments in two and three dimensions.

(Retracted) Measurement method and experimental study of complex profile of turbine blade based on image fusion

The hydraulic turbine is the core machinery of tidal power generation, and the performance of the hydraulic turbine determines the installed power and power generation efficiency. The purpose of this paper is to study the measurement method of turbine blades. A method for identifying the turbine blade based on image fusion technology is proposed. Based on the image recognition technology, the complex surface of the turbine blade is preprocessed, and then a method for extracting the complex surface of the turbine blade based on the genetic algorithm is proposed. This can effectively identify images based on image fusion processing, especially after the learning iteration of the genetic algorithm, and can achieve a good recognition effect. The experimental results of this paper show that the algorithm proposed in this paper has a measurement coverage rate of 99.70% for the turbine blades.

An iterative learning approach to formation control of discrete‐time multi‐agent systems With varying trial lengths

AbstractIn this article, an iterative learning approach is proposed for the formation control of discrete‐time multi‐agent systems, where the trial length of each learning iteration is randomly varying. In particular, a modified state error related to the prescribed formation is defined by taking into account the nonuniform actual trial length that could be different from the desired one. Then, a P‐type iterative learning protocol is established for switching networks of agents subject to nonuniform trial lengths, and the convergence analyses are given for the fixed and the iteration‐varying initial conditions respectively. It shows that through iterative learning, the given formation will be maintained among multiple agents in the entire time interval of one trial. In the end, simulations are done to demonstrate the correctness of the obtained theoretical results.

Classification of Skin Lesion through Active Learning Strategies

Background and objective: According to the National Cancer Institute, among all malignant tumors, non-melanoma skin cancer, and melanoma are the most frequent in Brazil. Despite having a lower incidence, the melanoma type has accelerated growth and greater lethality. Several studies have been performed in recent years in the computer vision area to assist in the early diagnosis of skin cancer. Despite being widely used and presenting good results, deep learning approaches require a large amount of annotated data and considerable computational cost for training the model. Therefore, the present work explores active learning approaches to select a small set of more informative data for training the classifier. For that, different selection criteria are considered to obtain more effective and efficient classifiers for skin lesions.Methods: We perform an extensive experimental evaluation considering three datasets and different learning strategies and scenarios for validation. In addition to data augmentation, we evaluated two segmentation strategies considering the U-net CNN model and the Fully Convolutional Networks (FCN) with a manual expert review. We also analyzed the best (handcrafted and deep) features that describe each skin lesion and the most suitable classifiers and combinations (extractor-classifier) for this context. The active learning approach evaluated different criteria based on uncertainty, diversity, and representativeness to select the most informative samples. The strategies used were Decreasing Boundary Edges, Entropy, Least Confidence, Margin Sampling, Minimum-Spanning Tree Boundary Edges, and Root-Distance based Sampling.Results: It can be observed that the segmentation with FCN and manual correction by the specialist, the Border-Interior Classification (BIC) extractor, and the Random Forest (RF) classifier showed a better performance. Regarding the active learning approach, the Margin Sampling strategy presented the best classification accuracies (about 93%) with only 35% of the training set compared to the traditional learning approach (which requires the entire set).Conclusions: According to the results, it is possible to observe that the selection strategies allow for achieving high accuracies faster (fewer learning iterations) and with a smaller amount of labeled samples compared to the traditional learning approach. Hence, active learning can contribute significantly to the diagnosis of skin lesions, beneficially reducing specialists’ annotation costs.

An AI-Based Adaptive Surrogate Modeling Method for the In-Service Response of UVLED Modules

The response forecasting of in-service complex electronic systems remains a challenge due to its uncertainty. An AI-based adaptive surrogate modeling method, including offline and online learning procedures, is proposed in this research for different systems with significant variety. The offline learning aims to abstract the knowledge from the known information and represent it as root models. The in-service response is modeled by a linear combination of the online learning of these root models against the continuous new measurement. This research applies a performance measurement dataset of the UVLED modules with considerable deviation to verify the proposed method. Part of the datasets is selected to generate the root models by offline learning, and these root models are applied to the online learning procedures for the adaptive surrogate model (ASM) of the different systems. The results show that after approximately 10 online learning iterations, the ASM achieves the capability of predicting 1000 h of response.