Advanced Learning Methods Research Articles

Deep reinforcement learning for real-world quadrupedal locomotion: a comprehensive review

Building controllers for legged robots with agility and intelligence has been one of the typical challenges in the pursuit of artificial intelligence (AI). As an important part of the AI field, deep reinforcement learning (DRL) can realize sequential decision making without physical modeling through end-to-end learning and has achieved a series of major breakthroughs in quadrupedal locomotion research. In this review article, we systematically organize and summarize relevant important literature, covering DRL algorithms from problem setting to advanced learning methods. These algorithms alleviate the specific problems encountered in the practical application of robots to a certain extent. We first elaborate on the general development trend in this field from several aspects, such as the DRL algorithms, simulation environments, and hardware platforms. Moreover, core components in the algorithm design, such as state and action spaces, reward functions, and solutions to reality gap problems, are highlighted and summarized. We further discuss open problems and propose promising future research directions to discover new areas of research.

Analysis of ground surface settlement in anisotropic clays using extreme gradient boosting and random forest regression models

Excessive ground surface settlement induced by pit excavation (i.e. braced excavation) can potentially result in damage to the nearby buildings and facilities. In this paper, extensive finite element analyses have been carried out to evaluate the effects of various structural, soil and geometric properties on the maximum ground surface settlement induced by braced excavation in anisotropic clays. The anisotropic soil properties considered include the plane strain shear strength ratio (i.e. the ratio of the passive undrained shear strength to the active one) and the unloading shear modulus ratio. Other parameters considered include the support system stiffness, the excavation width to excavation depth ratio, and the wall penetration depth to excavation depth ratio. Subsequently, the maximum ground surface settlement of a total of 1479 hypothetical cases were analyzed by various machine learning algorithms including the ensemble learning methods (extreme gradient boosting (XGBoost) and random forest regression (RFR) algorithms). The prediction models developed by the XGBoost and RFR are compared with that of two conventional regression methods, and the predictive accuracy of these models are assessed. This study aims to highlight the technical feasibility and applicability of advanced ensemble learning methods in geotechnical engineering practice.

Learning End-to-End Lossy Image Compression: A Benchmark.

Image compression is one of the most fundamental techniques and commonly used applications in the image and video processing field. Earlier methods built a well-designed pipeline, and efforts were made to improve all modules of the pipeline by handcrafted tuning. Later, tremendous contributions were made, especially when data-driven methods revitalized the domain with their excellent modeling capacities and flexibility in incorporating newly designed modules and constraints. Despite great progress, a systematic benchmark and comprehensive analysis of end-to-end learned image compression methods are lacking. In this paper, we first conduct a comprehensive literature survey of learned image compression methods. The literature is organized based on several aspects to jointly optimize the rate-distortion performance with a neural network, i.e., network architecture, entropy model and rate control. We describe milestones in cutting-edge learned image-compression methods, review a broad range of existing works, and provide insights into their historical development routes. With this survey, the main challenges of image compression methods are revealed, along with opportunities to address the related issues with recent advanced learning methods. This analysis provides an opportunity to take a further step towards higher-efficiency image compression. By introducing a coarse-to-fine hyperprior model for entropy estimation and signal reconstruction, we achieve improved rate-distortion performance, especially on high-resolution images. Extensive benchmark experiments demonstrate the superiority of our model in rate-distortion performance and time complexity on multi-core CPUs and GPUs.

Implementation of Mobile Entrepreneurial Learning in the Context of Flexible Integration of Traditions and Innovations

The study considers the issue of integrating traditions and principles of mobile learning in entrepreneurship education. The purpose of the research is to study the level of implementation of mobile learning methods in higher education through the example of the Republic of Azerbaijan. The study involves a survey of 99 teachers from 5 universities. The survey results confirmed that both external factors (primarily globalization and widespread introduction of information and communication technologies) and internal factors affect higher education. It has been shown that national traditions do not contradict the best practices of entrepreneurship education and are successfully integrated into mobile learning practices. The level of integration of advanced trends in the transformation of education in the practice of mobile learning has been highly assessed by the teachers (3.44 out of 5 points on the Likert scale). The benefit of mobile learning for implementing the identified advanced trends (3.44 - 4.84 on the Likert scale) has also been highly praised. For example, respect for elders by age, experience, rank; practical focus; comprehensive assessment; rationality; involvement of each family member in hard work are supported by advanced learning methods, including the involvement of skilled lecturers, the increased importance of practical tasks, the use of project approach and startups; the use of mobile applications to assess knowledge; personalization of classes; the introduction of distance learning, etc. Flexible integration of national traditions and innovations can become a competitive advantage of the university and the education system; it is also being successfully implemented in mobile learning. A model for the formation of entrepreneurial competence in students based on the use of mobile applications along with cloud services that takes into account the flexible integration of traditions and innovation has been developed.

2019 DAC Roundtable

As demands for computing have been continuously increasing, various solutions have been proposed. Some approaches deal with improving algorithms and software programs, mainly through the tuning of advanced heuristics and learning methods. Some other tackle the problem by providing specialized hardware to enhance computation. Other revolutionary approaches, such as memcomputing and quantum computing, explore new paradigms in computation to beat the barrier of computational complexity. A highly attended plenary panel at the 2019 Design Automation Conference (DAC) in Las Vegas, NV, USA, with the provocative title “Evolutionary Computing or Heuristic Forever?” spurred a lively discussion that is reported in this Roundtable. It is moderated by the panel organizer and moderator Giovanni De Micheli, EPFL, Switzerland. Panelists include Antun Domic, former CTO and senior VP at Synopsys; Jason Cong, University of California Los Angeles (UCLA); Massimiliano Di Ventra, University of California San Diego (UCSD); and Martin Roettler, Microsoft Research.

Learning Kernel-Based Robust Disturbance Dictionary for Face Recognition

In this paper, a kernel-based robust disturbance dictionary (KRDD) is proposed for face recognition that solves the problem in modern dictionary learning in which significant components of signal representation cannot be entirely covered. KRDD can effectively extract the principal components of the kernel by dimensionality reduction. KRDD not only performs well with occluded face data, but is also good at suppressing intraclass variation. KRDD learns the robust disturbance dictionaries by extracting and generating the diversity of comprehensive training samples generated by facial changes. In particular, a basic dictionary, a real disturbance dictionary, and a simulated disturbance dictionary are acquired to represent data from distinct subjects to fully represent commonality and disturbance. Two of the disturbance dictionaries are modeled by learning few kernel principal components of the disturbance changes, and then the corresponding dictionaries are obtained by kernel discriminant analysis (KDA) projection modeling. Finally, extended sparse representation classifier (SRC) is used for classification. In the experimental results, KRDD performance displays great advantages in recognition rate and computation time compared with many of the most advanced dictionary learning methods for face recognition.

Kernel Selection for Support Vector Machines for System Identification of a CNC Machining Center

Advanced learning methods enable the model-based control of systems with complex unknown dependencies. Within the German cluster of excellence “Internet of Production”, a configuration for an interconnected data-base is proposed, where data-driven model-based control strategies can be applied using the collective knowledge and adapted online according to data. For the exchanged data it is imperative to establish a generalizing learning technique for the controller design. A machine learning technique with inherent generalization ability is the Support Vector Machines (SVM) algorithm, where the choice of kernel is crucial for the resulting model quality. In the related literature, usually a radial basis function (RBF) is chosen as kernel, although many studies show the necessity of a more sophisticated kernel selection. This work tackles the point of a kernel selection based on composite kernel search in context of data-driven model-based control of a CNC machining center. The results support the capability of the presented approach to further automate and improve the identification of the controller model for the machining center.

Special Issue on Advanced Learning Methods for Complex Data

The importance of data mining methods has increased dramatically in recent years, making this research area relevant and challenging to extract actionable knowledge from complex data. Indeed, new algorithms and machine learning methods are constantly being explored to deal with domains that present multiple challenges including high-dimensionality, heterogeneity of features, and complex relationships between data objects. This special issue aims at discussing emerging approaches for learning from complex data, including text data, images, and social media data.

Helping Students Learn About Electromagnetics [Book/Software Reviews

The reviewer of the book Conceptual Electromagnetics notes that the book describes the fundamental subject of electromagnetics, or field and wave theory, for an undergraduate college class using very advanced teaching and learning methods. It provides abundant opportunities for instructors to introduce innovative lectures, in-class and homework assignments, and tests (including those for online distance education and for students pursuing independent learning). An active teaching and learning methodology is built into this text by means of interactive in-class questions, explorations, and discussions. The book also includes peer instruction, allowing students to tutor one another. The learner-centered pedagogies and practices - and active teaching and learning, in particular - are very effective, motivational, and positively evaluated by students. There are many conceptual questions as well as multiple-choice ones that focus on core concepts of the material, requiring conceptual reasoning and understanding. This book is organized into 12 chapters on electrostatics fields, steady electric currents, magnetostatic fields, time-varying electromagnetic fields, uniform plan electromagnetic waves, transmission lines, waveguides and cavity resonators, and antennas and wireless communication systems. For many engineering students, the fields and waves, or electromagnetics, course is particularly challenging. But, with the author' advanced teaching and learning approach, this course can be pleasant and fun. His teaching technique should become the standard for all science and engineering courses.

A Beta basis function Interval Type-2 Fuzzy Neural Network for time series applications

The huge complexity and uncertainty in real life requires the use of advanced automatic learning methods to find out better approximators and suitable relationship in real data behavior. Neuro fuzzy systems have been proved to be excellent universal approximators. In this paper we propose a new based function Interval Type-2 Fuzzy Neural Network denoted ”Beta basis function Interval Type-2 Fuzzy Neural Network”, the BIT2FNN. The main idea is to involve type-2 beta fuzzy sets in the design process of fuzzy networks. The proposed architecture is based on beta type-2 fuzzy sets in the antecedent part, while the consequent part achieves the TSK (Takagi–Sugeno–Kang) fuzzy output strategy. Thanks to the beta function flexibility, the network achieve a good performance and shows a good resistance to noisy data. First order derivatives of type-1 and type-2 Beta functions were developed for the first time for designing fuzzy logic systems based on given input–output pairs. The backpropagation algorithm was used for the learning process of antecedent fuzzy beta parameters and the consequent part. The performance of the proposed model of Beta fuzzy logic system is evaluated with mainly two problems of time series applications : the Mackey Glass Chaotic Time-Series prediction problem with different setting of parameters and levels of noise and the ECG heart-rate Time Series monitoring problem.

Simulation Training: Evaluating the Instructor's Contribution to a Wizard of Oz Simulator in Obstetrics and Gynecology Ultrasound Training.

BackgroundWorkplaces today demand graduates who are prepared with field-specific knowledge, advanced social skills, problem-solving skills, and integration capabilities. Meeting these goals with didactic learning (DL) is becoming increasingly difficult. Enhanced training methods that would better prepare tomorrow’s graduates must be more engaging and game-like, such as feedback based e-learning or simulation-based training, while saving time. Empirical evidence regarding the effectiveness of advanced learning methods is lacking. Objective quantitative research comparing advanced training methods with DL is sparse.ObjectivesThis quantitative study assessed the effectiveness of a computerized interactive simulator coupled with an instructor who monitored students’ progress and provided Web-based immediate feedback.MethodsA low-cost, globally accessible, telemedicine simulator, developed at the Technion—Israel Institute of Technology, Haifa, Israel—was used. A previous study in the field of interventional cardiology, evaluating the efficacy of the simulator to enhanced learning via knowledge exams, presented promising results of average scores varying from 94% after training and 54% before training (n=20) with P<.001. Two independent experiments involving obstetrics and gynecology (Ob-Gyn) physicians and senior ultrasound sonographers, with 32 subjects, were conducted using a new interactive concept of the WOZ (Wizard of OZ) simulator platform. The contribution of an instructor to learning outcomes was evaluated by comparing students’ knowledge before and after each interactive instructor-led session as well as after fully automated e-learning in the field of Ob-Gyn. Results from objective knowledge tests were analyzed using hypothesis testing and model fitting.ResultsA significant advantage (P=.01) was found in favor of the WOZ training approach. Content type and training audience were not significant.ConclusionsThis study evaluated the contribution of an integrated teaching environment using a computerized interactive simulator, with an instructor providing immediate Web-based immediate feedback to trainees. Involvement of an instructor in the simulation-based training process provided better learning outcomes that varied training content and trainee populations did not affect the overall learning gains.

イノベーション人材育成のための社会連携型教育における高次の学習法

Today’ s postmodern society especially has higher expectations of engineers to acquire abilities as innovative human resources. Innovation education, a new genre of education for engineers, is such a learning process that goes beyond the range of knowledge and experiences provided by traditional engineering education. To develop engineers with these required abilities, one type of active learning that experiences the process called Social Implementation has proved effective. Accordingly, we have designed numerous educational activities and implemented them in the learning process of imparting technology to society. Trial and subsequent improvements over nine years has revealed that information management skills, discussion and consensus-building skills, and evaluation skills and the learning methods to acquire those skills are critical. This paper introduces advanced learning methods based on situated cognition theories and the tools to further develop the quality of active learning for engineering students who strive to be innovators.

Advanced learning methods and exponent regularization applied to a high order neural network

High order neural networks (HONNs) are neural networks which employ neurons that combine their inputs non-linearly. The high order network with exponential synaptic links (HONEST) network is a HONN that uses neurons with product units and adaptable exponents. This study examines the use of several advanced learning methods to train the HONEST network: resilient propagation, conjugate gradient, scaled conjugate gradient (SCG), and the Levenberg---Marquardt method. Using a collection of 32 widely-used benchmark datasets, we compare the mean squared error (MSE) performance of the HONEST network across the four algorithms, in addition to backpropagation, and find the SCG method to produce the best performance to a statistically significant extent. Additionally, we investigate the use of a regularization term in the error function, to smooth the magnitudes of the network exponents and nudge the network towards smaller exponents. We find that the use of regularization reduces exponent magnitudes without compromising test set MSE performance.

Policy analytics: an agenda for research and practice

The growing impact of the “analytics” perspective in recent years, which integrates advanced data-mining and learning methods, is often associated with increasing access to large databases and with decision support systems. Since its origin, the field of analytics has been strongly business-oriented, with a typical focus on data-driven decision processes. In public decisions, however, issues such as individual and social values, culture and public engagement are more important and, to a large extent, characterise the policy cycle of design, testing, implementation, evaluation and review of public policies. Therefore public policy making seems to be a much more socially complex process than has hitherto been considered by most analytics methods and applications. In this paper, we thus suggest a framework for the use of analytics in supporting the policy cycle—and conceptualise it as “Policy Analytics”.

A Web-Based Multimedia Retrieval System with MCA-Based Filtering and Subspace-Based Learning Algorithms

The popularity of research on intelligent multimedia services and applications is motivated by the high demand of the convenient access and distribution of pervasive multimedia data. Facing with abundant multimedia resources but inefficient and rather old-fashioned keyword-based retrieval approaches, Intelligent Multimedia Systems (IMS) demand on (i) effective filtering algorithms for storage saving, computation reduction, and dynamic media delivery; and (ii) advanced learning methods to accurately identify target concepts, effectively search personalized media content, and enable media-type driven applications. Nowadays, the web based multimedia applications become more and more popular. Therefore, how to utilize the web technology into multimedia data management and retrieval becomes an important research topic. In this paper, the authors developed a web-based intelligent video retrieval system that integrates effective and efficient MCA-based filtering and subspace-based learning to facilitate end users to retrieve their desired semantic concepts. A web-based demo shows the effectiveness of the proposed intelligent multimedia system to provide relevant results of target semantic concepts retrieved from TRECVID video collections.

Failure and reliability prediction by support vector machines regression of time series data

Support Vector Machines (SVMs) are kernel-based learning methods, which have been successfully adopted for regression problems. However, their use in reliability applications has not been widely explored. In this paper, a comparative analysis is presented in order to evaluate the SVM effectiveness in forecasting time-to-failure and reliability of engineered components based on time series data. The performance on literature case studies of SVM regression is measured against other advanced learning methods such as the Radial Basis Function, the traditional MultiLayer Perceptron model, Box-Jenkins autoregressive-integrated-moving average and the Infinite Impulse Response Locally Recurrent Neural Networks. The comparison shows that in the analyzed cases, SVM outperforms or is comparable to other techniques.

Gradient Descent Learning for Rotor Associative Memory

Complex-valued Associative Memory (CAM) is an extended model of Hopfield Associative Memory (HAM). The fundamental elements, such as input-output signals and connection weights of the CAM are extended to complex numbers. The CAM can deal with multi-states information. Rotor Associative Memory (RAM) is an extended model of the CAM. Rotor neurons are essentially equivalent to complex-valued neurons. Connection weights of the RAM are expressed by two by two matrices. Only hebb rule has been proposed for the learning of the RAM. Its storage capacity is small, so advanced learning methods are necessary. In this paper, we propose gradient descent learning rule for the RAM (GDR RAM). It is based on that for the CAM (GDR CAM) proposed by Lee. We solved the learning rule and performed computer simulations to compare the GDR CAM and the GDR RAM. At last, it turned out that the storage capacity of the GDR RAM is approximately twice as much as that of the GDR CAM and the noise robustness of the GDR RAM is much better than that of the GDR CAM.

Understanding Crash Mechanisms and Selecting Interventions to Mitigate Real-Time Hazards on Urban Expressways

The concept of dynamic black spot identification has matured substantially over the past decade. Rather than being confined to improving prediction accuracy, the concept has expanded to understanding the crash mechanism by using sensor data and devising real-time countermeasures. A major drawback of most studies has been considering a crash as a generic mechanism throughout the expressway. Consequently, proposed countermeasures, such as warning messages, variable speed limits, and ramp metering, have been independent of location. Some studies have incorporated ramps as variables, but little progress has been made in investigating variation in the crash mechanism or evaluating the effectiveness of countermeasures separately for the basic freeway segments and ramp vicinities. This paper divides expressways into five segments: basic freeway, upstream and downstream of exits, and entrance ramps. The paper uses high-resolution detector data and advanced ensemble learning methods such as random forest and classification and regression trees to identify important factors and understand the crash mechanism. The outcome advocates lane-based speed management to improve the safety of basic freeway segments. Although ramp flows were found to contribute to crashes in ramp vicinities, the flows caused little concern for basic freeway segments. The findings will be valuable for researchers in predicting real-time crashes, investigating crash mechanisms, and designing proactive countermeasures.

Modified cascade-correlation learning for classification

The main advantages of cascade-correlation learning are the abilities to learn quickly and to determine the network size. However, recent studies have shown that in many problems the generalization performance of a cascade-correlation trained network may not be quite optimal. Moreover, to reach a certain performance level, a larger network may be required than with other training methods. Recent advances in statistical learning theory emphasize the importance of a learning method to be able to learn optimal hyperplanes. This has led to advanced learning methods, which have demonstrated substantial performance improvements. Based on these recent advances in statistical learning theory, we introduce modifications to the standard cascade-correlation learning that take into account the optimal hyperplane constraints. Experimental results demonstrate that with modified cascade correlation, considerable performance gains are obtained compared to the standard cascade-correlation learning. This includes better generalization, smaller network size, and faster learning.

Reading to Learn: A Project in Advanced Learning Methods

Reading to Learn: A Project in Advanced Learning Methods