Stable Learning Research Articles

Review Seminar Based on CRS in Introductory Physics

The research on the relationship between the teaching model of seminar and the learning gain of students has been a hot topic for a long time. As the Classroom Response System (CRS) is widely used in interactive teaching environments to attract students by asking questions. This paper takes electromagnetism of introductory physics as an example to study the teaching mode of review seminar based on CRS. Focusing on the necessity of review seminar based on CRS, the teaching efficiency as well as the stability of students' learning effect under this teaching mode, this paper qualitatively discusses the necessity and teaching efficiency of review seminar based on CRS, and quantifies the learning effect of students under this mode from the correct transformation of students' concept and its stability. The results show that the teaching model of seminar based on CRS is suitable for the teaching of problem sets in introductory and has high teaching efficiency. Most of the students have the right change of physics concept after the discussion with their peers, and have stable learning gain. At the end of the paper, we discuss the shortcomings of the research and looks forward to the future research direction.

On the Design of Tailored Neural Networks for Energy Harvesting Broadcast Channels: A Reinforcement Learning Approach

In this paper, we consider a power allocation optimization technique for a time-varying fading broadcast channel in energy harvesting communication systems, in which a transmitter with a rechargeable battery transmits messages to receivers using the harvested energy. We first prove that the optimal online power allocation policy for the sum-rate maximization of the transmitter is a monotonically increasing function of harvested energy, remaining battery, and the channel gain of each user. We then construct a lightweight neural network architecture to take advantage of the monotonicity of the optimal policy. This two-step approach, which relies on effective function approximation to provide a mathematical guideline for neural network design, can prevent us from wasting the representational capacity of neural networks. The tailored neural network architectures enable stable learning and eliminate the heuristic network design. For performance assessment, the proposed approach is compared with the closed-form optimal policy for a partially observable Markov problem. Through additional experiments, it is observed that our online solution achieves a performance close to the theoretical upper bound of the performance in a time-varying fading broadcast channel.

Dual-Branch Structured De-Striping Convolution Network Using Parametric Noise Model

The stripe fixed pattern noise (FPN) of infrared images significantly corrupts image quality, so that most infrared imaging systems suffer from the degradation of visibility and detectability during operation. Therefore, the FPN de-striping method, which eliminates stripe patterns without substantial loss of image information, remains a core technology in the field of infrared image processing. In this article, we propose the dual-branch structure based FPN de-striping deep convolutional neural network (DBS-DCN) to effectively extract structural features of FPN and preserve the image details in a single infrared image. In addition, we have established the parametric FPN model through the diagnostic experiments of infrared images based on the physical principle of an infrared detector and its signal response. We have optimized each parameter of the FPN model using measured data, which acquired on a wide range of detector temperatures. Further, we generate the training data using our FPN model to ensure stable learning performance against various stripe patterns. We performed comparative experiments with state-of-the-art methods using artificially corrupted infrared images and real corrupted infrared data, and our proposed method achieved outstanding de-striping results in both qualitative and quantitative evaluation compared to existing methods.

Novel D-SLP Controller Design for Nonlinear Feedback Control

Novel nonlinear feedback control based on the dragonfly swarm learning process (D-SLP) algorithm is proposed in this paper. This approach improves the performance, stability and robustness of designing the nonlinear system controller. The D-SLP algorithm is the combination of the dragonfly algorithm (DA) and swarm learning process (SLP) algorithm by applying the DA to the learning process of the SLP algorithm. Furthermore, the estimation of the nonlinear term by using gradient descent is proposed in the process of the D-SLP algorithm. The learning rate is adjusted according to the stable learning rate, which is derived according to the Lyapunov stability theorem. To show the superior performance and robustness of the proposed control method, it is compared with the simulation of designing the controller based on a permanent magnet synchronous motor (PMSM) control system with the online autotuning parameter of a PID controller and LQR controller with two case studies. The conventional SLP algorithm and DA are used to autotune the PID controller, while an artificial bee colony algorithm and a flower pollination algorithm (ABC-FPA) autotune the LQR controller. From the simulation results, the proposed control method can provide a better response than the other control method. Additionally, the global convergence of the D-SLP algorithm is analyzed according to Markov chain modeling and proved to correspond with the policy of global convergence for stochastic search algorithms.

Video Object Segmentation by Latent Outcome Regression

This paper presents a novel algorithm for unsupervised video object segmentation (UVOS) in unconstrained scenarios. Although a large variety of methods have been proposed in the literature, segmenting generic objects is still challenging because different methods often perform well in different situations, and no single method can outperform the others in all cases. To address this, we propose to solve the problem of UVOS in a crowd-sourcing setting. We claim that one can achieve superior results by aggregating the predictions of multiple imperfect methods in a reasonable way. Specifically, we propose a latent regression algorithm for ensemble-based segmentation by jointly labelling pixels in a sequence and learning an adaptive weight for each single method in an ensemble. The pixel labellings offer the outcome (pseudo groundtruth) for regression and thus promote the procedure of weight learning, while the learnt weights could provide better shape priors for labelling, resulting in more accurate segmentation. Besides, Laplacian regularization is introduced into the regression to facilitate a stable learning of the weights. The most distinct feature of our algorithm is that it adaptively learns the contributions of different single methods for each test sequence, thus is capable of capturing the advantages of those methods while avoiding their weaknesses. In the experiments, our algorithm is built on 14 non-deep learning segmentation methods which are based on handcrafted features and require no training data. Experimental results on popular benchmarks show that our algorithm achieves compelling performance, even in comparison with deep learning-based methods. Furthermore, benefiting from the adaptive weight learning mechanism, our algorithm can achieve good flexibility and usability by choosing the most complementary single methods without losing too much performance.

Simulation faculty and staff development: An interprofessional, online approach

Abstract Two academic healthcare programs and a university-affiliated hospital were faced with increasing demands for faculty and staff knowledge in support of expanded use of simulation-based education. While a few authors have published recommendations and innovations for faculty development in simulation programs, 1 best practices have yet to be determined. 2 , 3 Through collaboration by simulation program administrators and educators at the College of Nursing, the School of Medicine, and the hospital, an online program was developed for use by multiple disciplines and facilities. An online approach allows for standardization of simulation faculty development across three entities. Faculty and staff were incentivized by the ability to receive continuing education credits. The online self-directed program includes 10 modules with a learner pretest and posttest. Comparisons of pre- and post-test scores have shown knowledge improvements. Through post-course surveys, learners have indicated favorable responses to the online program. In order to implement this type of online approach to simulation faculty development, a stable web-based learning platform is required. For more information about the course, the lead author may be contacted at XXXX. Format The online simulation educator course consists of 10 modules with pretest, posttest, and module quizzes for 8 of 10 modules. Within the program, learners are asked about their personal objectives at the beginning and whether they accomplished those at the end of the course. The Eastern Area Health Education Center (Eastern AHEC), which hosts the program on its learning platform, requires a learner evaluation to be completed before issuing the continuing education certificate, to assure that the data is obtained.

Constrained Deep Q-Learning Gradually Approaching Ordinary Q-Learning.

A deep Q network (DQN) (Mnih et al., 2013) is an extension of Q learning, which is a typical deep reinforcement learning method. In DQN, a Q function expresses all action values under all states, and it is approximated using a convolutional neural network. Using the approximated Q function, an optimal policy can be derived. In DQN, a target network, which calculates a target value and is updated by the Q function at regular intervals, is introduced to stabilize the learning process. A less frequent updates of the target network would result in a more stable learning process. However, because the target value is not propagated unless the target network is updated, DQN usually requires a large number of samples. In this study, we proposed Constrained DQN that uses the difference between the outputs of the Q function and the target network as a constraint on the target value. Constrained DQN updates parameters conservatively when the difference between the outputs of the Q function and the target network is large, and it updates them aggressively when this difference is small. In the proposed method, as learning progresses, the number of times that the constraints are activated decreases. Consequently, the update method gradually approaches conventional Q learning. We found that Constrained DQN converges with a smaller training dataset than in the case of DQN and that it is robust against changes in the update frequency of the target network and settings of a certain parameter of the optimizer. Although Constrained DQN alone does not show better performance in comparison to integrated approaches nor distributed methods, experimental results show that Constrained DQN can be used as an additional components to those methods.

Indirect Neural Adaptive Control for Wheeled Mobile Robot

For a precise trajectory tracking of a wheeled mobile robot, accurate control of the position along a reference trajectory is essential. Therefore, this paper proposes an indirect neural adaptive controller for a nonholonomic mobile robot based on its dynamical model. This controller takes into account the approximation error. The use of the Lyapunov stability theorem and dynamical neural networks is indeed for deriving respectively stable learning laws for control and identification of a complex nonlinear dynamics system. The global tracking error is incorporated to adjust the neural weight learning laws to ensure the robustness of the system against approximation inaccuracy. Hence, the designed intelligent controller guarantees the convergence of both tracking and identification errors to zero. Simulation results illustrate the ability of the intelligent controller to assure the asymptotic stability of the closed-loop nonlinear uncertain system.

Multidimensional Taylor network adaptive control for MIMO time‐varying uncertain nonlinear systems with noises

SummaryIn this paper, an adaptive control approach based on the multidimensional Taylor network (MTN) is proposed here for the real‐time tracking control of multiple‐input–multiple‐output (MIMO) time‐varying uncertain nonlinear systems with noises. Two MTNs are used to formulate the optimum control and adaptive filtering approaches. The feed‐forward MTN controller (MTNC) is developed to realize the precise tracking control. The closed‐loop errors between the filtered outputs and expected values are directly chosen as the MTNC's inputs. A valid initial value selection scheme for the weights of the MTNC, which can ensure the initial stability of adaptive process, is introduced. The proposed MTNC can update its weights online according to errors caused by system's uncertain factors, based on stable learning rate. The resilient backpropagation algorithm and the adaptive variable step size algorithm via linear reinforcement are utilized to update the MTNC's weights. The MTN filter (MTNF) is developed to eliminate measurement noises and other stochastic factors. The proposed adaptive MTN filtering system possesses the distinctive properties of the Lyapunov theory–based adaptive filtering system and MTN. Lyapunov function of the filtering errors between the measured values and MTNF's outputs is defined. By properly choosing the weights update law in the Lyapunov sense, the MTNF's outputs can asymptotically converge to the desired signals. The design is independent of the stochastic properties of the input disturbances. Simulation of the MTN‐based control is conducted to test the effectiveness of the presented results.

Uncertainty-based modulation for lifelong learning

The creation of machine learning algorithms for intelligent agents capable of continuous, lifelong learning is a critical objective for algorithms being deployed on real-life systems in dynamic environments. Here we present an algorithm inspired by neuromodulatory mechanisms in the human brain that integrates and expands upon Stephen Grossberg’s ground-breaking Adaptive Resonance Theory proposals. Specifically, it builds on the concept of uncertainty, and employs a series of “neuromodulatory” mechanisms to enable continuous learning, including self-supervised and one-shot learning. Algorithm components were evaluated in a series of benchmark experiments that demonstrate stable learning without catastrophic forgetting. We also demonstrate the critical role of developing these systems in a closed-loop manner where the environment and the agent’s behaviors constrain and guide the learning process. To this end, we integrated the algorithm into an embodied simulated drone agent. The experiments show that the algorithm is capable of continuous learning of new tasks and under changed conditions with high classification accuracy (>94%) in a virtual environment, without catastrophic forgetting. The algorithm accepts high dimensional inputs from any state-of-the-art detection and feature extraction algorithms, making it a flexible addition to existing systems. We also describe future development efforts focused on imbuing the algorithm with mechanisms to seek out new knowledge as well as employ a broader range of neuromodulatory processes.

Preventing interference: Reordering complexity in the learning of new concepts.

Introducing new concepts to learners in an order of increasing complexity appears to be beneficial for learning, but typically introduction of concepts does not always adhere to this principle. We examined whether introducing new algebra concepts in a contrasted manner or in an order of increasing complexity instead of a different more typical order influences learning and retention. A total of 138 sixth graders were introduced to algebraic expression simplification. In algebra, addition of terms is usually introduced before multiplication of terms. However, the concept of addition appears to be structurally more complex than the concept of multiplication. Two experimental conditions in which algebraic multiplication was either introduced first, or both concepts simultaneously, caused substantially stronger learning gains and better retention compared with a third condition with the more typical order. Contrasted presentation of both concepts performed best, with strongest and most stable learning gains. Presenting multiplication first led to comparable gains on addition problems, but to lower achievement on multiplication problems in the longer delay. These findings suggest a rethinking of introduction orders in algebra and beyond. (PsycINFO Database Record (c) 2019 APA, all rights reserved)

Error-Based Learning Mechanism for Fast Online Adaptation in Robot Motor Control.

Existing state-of-the-art frequency adaptation mechanisms of central pattern generators (CPGs) for robot locomotion control typically rely on correlation-based learning. They do not account for the tracking error that may occur between the actual system motion and CPG output, leading to the loss of precision, unwanted movement, inefficient energy locomotion, and in the worst cases, motor collapse. To overcome this problem, we developed online error-based learning for frequency adaptation of CPGs. The learning mechanism used for error reduction is a novel modification of the dual learner (DL) called dual integral learner (DIL). Being able to reduce tracking and steady-state errors, it can also perform fast and stable learning, adapting the CPG frequency to match the performance of robotic systems. Control parameters of the DIL are more straightforward for complex systems (like walking robots), compared to traditional correlation-based learning, since they correspond to error reduction. Due to its embedded memory, the DIL can relearn quickly and recover spontaneously from the previously learned parameters. All these features are not covered by the existing frequency adaptation mechanisms. We integrated the DIL into a neural CPG-based motor control system for use on different legged robots with various morphologies for evaluation. The results show that: 1) the DIL does not require precise adjustment of its parameters to fit specific robots; and 2) the DIL can automatically and quickly adapt the CPG frequency to the robots such that the entire trajectory of the CPG can be precisely followed with very low tracking and steady-state errors. Consequently, the robots can perform the desired movements with more energy-efficient locomotion compared to the state-of-the-art correlation-based learning mechanism called frequency adaptation through fast dynamical coupling (AFDC). In the future, the proposed error-based learning mechanism for fast online adaptation in robot motor control can be used as a basis for trajectory optimization, universal controllers, and other studies concerning the change of intrinsic or extrinsic parameters.

Identification of nonlinear systems using adaptive variable-order fractional neural networks (Case study: A wind turbine with practical results)

In this paper, a Variable-Order Fractional Single-layer Neural Network (VOFSNN) and a Variable-Order Fractional Multi-layer Neural Network (VOFMNN) are proposed to identify nonlinear systems assuming all the system states are measurable. Fractional Lyapunov-like approach and Gronwall–Bellman integral inequality are employed to prove stability and asymptotic stability conditions of the identification error dynamics. A set of novel stable learning rules for the fractional order, the hidden layer weights and the output layer weights are derived to update the proposed VOFSNN and VOFMNN parameters. The proposed methods capabilities are evaluated and confirmed by the practical data gathered from a wind turbine under operation in a wind farm.

Improved hybrid recommendation with user similarity for adult learners

Nowadays, adult learners like to study diverse and personal materials in e-learning of continuing education. Even though some materials are conducted by fewer learners, they can still be propagated through the whole learning community. Meanwhile, the huge online materials cannot meet learners’ requirements. It leads to most learners encountering the e-learning problems of ‘resource overload’ or ‘learning loses’, and learners giving up studying easily. So this study introduces an improved hybrid recommendation with user similarity (IHUS) for adult learners, which can generate the user's list of importance based on the greatest user similarity. In IHUS, when the online system runs from cold starting, the tags calculation is conducted. When this system achieves a stable learning community, an improved leaderRank method is conducted.

Educational Technology for Active Connections in Blended Learning Environments

This conference focuses on employing educational technology for connections. It assumes that these connections will be active, and not latent – otherwise these will be meaningless and ineffective. The emphasis is thus on creating effective learning environments. Such learning environments are not just online or digital, but can also be physical learning spaces in which educational technology can play a key role.A key strategy to ensure that educational technology connections are indeed active is to employ educational technology within an active learning framework for both online and on-campus learning i.e. blended learning. Educational technology on its own does not lead to active learning – only when it is used within well-founded learning designs – of which constructive alignment is highlighted below. I am thus agreeing with Veletsianos and Moe (2017) that educational technology by itself is not “education’s silver bullet” but should be located within “the essentials of teaching and learning: theory, pedagogy and emergent trends in the research.” Such active learning will lead to learner engagement, leading to effective learning and learner success. The vision for instance for educational technologies at Charles Sturt University is to “support educational practices focussed on student success by providing cutting edge, stable learning environments”.

Fast and Stable Learning of Dynamical Systems Based on Extreme Learning Machine

The approach of dynamical system (DS) is promising for modeling robot motion, and provides a flexible means of realizing robot learning and control. Accuracy, stability, and learning speed are the three main factors to be considered when learning robot movements from human demonstrations with DS. Some approaches yield stable dynamical systems, but these may result in a poor reproduction performance, while some approaches yield good reproduction performance but are quite complex and time-consuming. In this paper, we address the accuracy-stability-speed issues simultaneously. We present a learning method named the fast and stable modeling for dynamical systems, which is based on the extreme learning machine to efficiently and accurately learn the parameters of the DS as well as to ensure the asymptotic stability at the target. We confirm the proposed approach by performing both 2-D tasks of learning handwriting motions and a set of robot experiments.

Forecasting Economy-Related Data Utilizing Weight-Constrained Recurrent Neural Networks

During the last few decades, machine learning has constituted a significant tool in extracting useful knowledge from economic data for assisting decision-making. In this work, we evaluate the performance of weight-constrained recurrent neural networks in forecasting economic classification problems. These networks are efficiently trained with a recently-proposed training algorithm, which has two major advantages. Firstly, it exploits the numerical efficiency and very low memory requirements of the limited memory BFGS matrices; secondly, it utilizes a gradient-projection strategy for handling the bounds on the weights. The reported numerical experiments present the classification accuracy of the proposed model, providing empirical evidence that the application of the bounds on the weights of the recurrent neural network provides more stable and reliable learning.

The consistency and cognitive predictors of children's oral language, reading, and math learning profiles

Efforts to understand learning disorders in children by focusing on specific domains and within restricted ability ranges have failed to identify consistent and stable learning profiles. Given evidence for dimensional distribution of oral language, reading, and mathematical skills among those identified with and without learning disorders, examining learning across a range of abilities and domains should provide a better estimate of learning profiles. The present study examined the 1-year stability of cross-domain learning profiles and associated cognitive characteristics of 327 children. Results revealed highly stable profiles with 95% of participants remaining in the same learning profiles across data years. Generally similar performance across domains was observed for three profiles (below average, average, above average) comprising 63% of the sample, with relatively specific differences in oral language or reading characterizing the remaining profiles. Cognitive measures and teacher ratings accurately predicted learning profile in about 55% of participants either at the time of testing or in the following year. The most effective models for categorizing learning profiles all included teacher ratings of reading. Cognitive measures of verbal working memory, verbal intelligence, phonological awareness, symbolic comparison, and visuospatial working memory were also important contributors to classification. The findings indicate that examining learning across domains, abilities, and time has the potential to inform our conceptualization of learning disorders and associated cognitive strengths and weaknesses.

Online and Stable Learning of Analysis Operators

In this paper, four iterative algorithms for learning analysis operators are presented. They are built upon the same optimization principle underlying both Analysis K-SVD and Analysis SimCO. The forward and sequential analysis operator learning (AOL) algorithms are based on projected gradient descent with optimally chosen step size. The implicit AOL algorithm is inspired by the implicit Euler scheme for solving ordinary differential equations and does not require to choose a step size. The fourth algorithm, singular value AOL, uses a similar strategy as Analysis K-SVD while avoiding its high computational cost. All algorithms are proven to decrease or preserve the target function in each step and a characterization of their stationary points is provided. Further they are tested on synthetic and image data, compared to Analysis SimCO and found to give better recovery rates and faster decay of the objective function, respectively. In a final denoising experiment the presented algorithms are again shown to perform similar to or better than the state-of-the-art algorithm ASimCO.

An Adaptive Control Algorithm for Stable Training of Generative Adversarial Networks

Generative adversarial networks (GANs) have shown significant progress in generating highquality visual samples, however they are still well known both for being unstable to train and for the problem of mode collapse, particularly when trained on data collections containing a diverse set of visual objects. In this paper, we propose an Adaptive k-step Generative Adversarial Network (Ak-GAN), which is designed to mitigate the impact of instability and saturation in the original by dynamically adjusting the ratio of the training steps of both the generator and discriminator. To accomplish this, we track and analyze stable training curves of relatively narrow datasets and use them as the target fitting lines when training more diverse data collections. Furthermore, we conduct experiments on the proposed procedure using several optimization techniques (e.g., supervised guiding from previous stable learning curves with and without momentum) and compare their performance with that of state-of-the-art models on the task of image synthesis from datasets consisting of diverse images. Empirical results demonstrate that Ak-GAN works well in practice and exhibits more stable behavior than regular GANs during training. A quantitative evaluation has been conducted on the Inception Score (IS) and the relative inverse Inception Score (RIS); compared with regular GANs, the former has been improved by 61% and 83%, and the latter by 21% and 60%, on the CelebA and the Anime datasets, respectively.