Input Weights Research Articles

Visual cues, expectations, and sensorimotor memories in the prediction and perception of object dynamics during manipulation

When we grasp and lift novel objects, we rely on visual cues and sensorimotor memories to predictively scale our finger forces and exert compensatory torques according to object properties. Recently, it was shown that object appearance, previous force scaling errors, and previous torque compensation errors strongly impact our percept. However, the influence of visual geometric cues on the perception of object torques and weights in a grasp to lift task is poorly understood. Moreover, little is known about how visual cues, prior expectations, sensory feedback, and sensorimotor memories are integrated for anticipatory torque control and object perception. Here, 12 young and 12 elderly participants repeatedly grasped and lifted an object while trying to prevent object tilt. Before each trial, we randomly repositioned both the object handle, providing a geometric cue on the upcoming torque, as well as a hidden weight, adding an unforeseeable torque variation. Before lifting, subjects indicated their torque expectations, as well as reporting their experience of torque and weight after each lift. Mixed-effect multiple regression models showed that visual shape cues governed anticipatory torque compensation, whereas sensorimotor memories played less of a role. In contrast, the external torque and committed compensation errors at lift-off mainly determined how object torques and weight were perceived. The modest effect of handle position differed for torque and weight perception. Explicit torque expectations were also correlated with anticipatory torque compensation and torque perception. Our main findings generalized across both age groups. Our results suggest distinct weighting of inputs for action and perception according to reliability.

SK2 channels in cerebellar Purkinje cells contribute to excitability modulation in motor-learning-specific memory traces.

Neurons store information by changing synaptic input weights. In addition, they can adjust their membrane excitability to alter spike output. Here, we demonstrate a role of such “intrinsic plasticity” in behavioral learning in a mouse model that allows us to detect specific consequences of absent excitability modulation. Mice with a Purkinje-cell–specific knockout (KO) of the calcium-activated K+ channel SK2 (L7-SK2) show intact vestibulo-ocular reflex (VOR) gain adaptation but impaired eyeblink conditioning (EBC), which relies on the ability to establish associations between stimuli, with the eyelid closure itself depending on a transient suppression of spike firing. In these mice, the intrinsic plasticity of Purkinje cells is prevented without affecting long-term depression or potentiation at their parallel fiber (PF) input. In contrast to the typical spike pattern of EBC-supporting zebrin-negative Purkinje cells, L7-SK2 neurons show reduced background spiking but enhanced excitability. Thus, SK2 plasticity and excitability modulation are essential for specific forms of motor learning.

Input Determination for Models Used in Predicting Student Performance

INTRODUCTION Student performance prediction has become a viable means to improving academic performance and course content in online learning. Predictive models such as neural networks, decision trees and linear regression are used to transform inputs (e.g. past performance, social background, learning system usage patterns, test results) into outputs (course completion, expected grade, difficulties encountered, personalized suggestions). Often, the existing quantitative data drive model design, especially when applying such models to the conventional classroom and the person delivering the course, is a passive participant in designing models and delivering data. In seeking to capture and code as much student behavior and environment as possible to apply learning analytics to a mostly conventional classroom, the most successful inputs (predictors) among existing models can be identified, categorized and their common characteristics determined. Together with a study of formative and summative assessment methods (e.g. types of feedback and how it can be captured) and factors affecting student performance in the classroom (e.g. environmental factors), this allows to identify the existing data in classrooms that are not captured by current learning management systems, thus allowing the expanded use of learning analytics and student performance prediction in traditional classrooms, with a focus on personalized suggestions. The goal of the paper is to identify patterns among inputs used in existing models of student learning (based on online learning and learning management system data mining) that can then also be applied to the traditional classroom. Research question: how can characteristics common to effective predictors of student performance be used to identify predictors among data produced in the traditional classroom? MATERIAL AND METHODS A literature review is performed where inputs captured and features discovered in existing learning analytics systems are characterised, along with methods used to identify those and the modelling approaches employed. An attempt is made to identify measures in online learning that may have analogues in the traditional classroom (e.g., seating patterns and communication in chatrooms) or for which proxies may be found (e.g. screen size and lighting quality, where the proxy is the classroom number). The corresponding outputs are recorded where possible, with a focus on those that allow providing feedback for individual students or for course/curriculum deliverers/designers (i.e. allow to improve the success of future students in this course). RESULTS Successful predictors and characteristics common to those are identified, so that they can be used in features engineering for student performance prediction models. Predictors used in online learning are categorised, so that analogous inputs can be developed for use in traditional classrooms. Types of feedback provided by existing models of learning are identified, where possible, along with the corresponding input (weights of inputs). Studies are identified where learning personnel, not the researcher, were able to drive the model development process. DISCUSSION Recently, there has been increasing focus on increasing the visibility into models of learning and of involving learning personnel in designing, modifying and running those models. Providing inputs and recognizing the features they represent determines the success of such models. Therefore, recognizing existing successes and applying them to formative assessment methods may be a means of recognizing additional inputs to and features used in models, while involving educators. Applying learning models to the traditional classroom as an integrated part of the learning management (school record keeping/grading) systems may allow to expand their use, while simultaneously increasing the predictive power and effectiveness of (personalized) suggestions, both by using existing data, and by providing tools for educators to transform the existing feedback they provide into data than can be used as inputs for models. CONCLUSION Predictors used in learning models in online learning can be applied to the traditional classroom. Analogues may be found for predictors that are not available in the conventional classroom. Common characteristics and categorisation of predictors may be used to identify predictors among existing data, including data provided by students (e.g. formative feedback) that is not captured by the existing learning management systems used.

A New Workload Prediction Model Using Extreme Learning Machine and Enhanced Tug of War optimization

Abstract Time series data is widely accessible in many life areas like economy, weather, stock price, retail sales, distributed system workloads. While many studies have focused on improving existing prediction techniques on accuracy aspect, less efforts is paid towards simple but efficient forecasting models in order to keep the balance between computation cost and prediction accuracy. In this work, we propose a novel time series prediction model, which aims to both model simplicity and accuracy. The core of the model is built based on extreme learning machine. Due to the random determination process for input weights and hidden biases, extreme learning machine requires a large number of hidden neurons to achieve good results and this increases the model complexity. To overcome this drawback, we propose a new opposition-based tug of war optimization to select optimally input weights, and hidden biases then apply to extreme learning machine. Two real public traffic monitoring datasets from Internet service providers were employed to evaluate our design. The achieved outcomes demonstrate that our proposed solution works effectively with satisfied performance in comparison with existing models for distributed system workload prediction.

ELM-Based Superimposed CSI Feedback for FDD Massive MIMO System

In frequency-division duplexing (FDD) massive multiple-input multiple-output (MIMO), deep learning (DL)-based superimposed channel state information (CSI) feedback has presented promising performance. However, it is still facing many challenges, such as the high complexity of parameter tuning, large number of training parameters, and long training time, etc. To overcome these challenges, an extreme learning machine (ELM)-based superimposed CSI feedback is proposed in this paper, in which the downlink CSI is spread and then superimposed on uplink user data sequence (UL-US) to feed back to base station (BS). At the BS, an ELM-based network is constructed to recover both downlink CSI and UL-US. In the constructed ELM-based network, we employ the simplified versions of ELM-based subnets to replace the subnets of DL-based superimposed feedback, yielding less training parameters. Besides, the input weights and hidden biases of each ELM-based subnet are loaded from the same matrix by using its full or partial entries, which significantly reduces the memory requirement. With similar or better recovery performances of downlink CSI and UL-US, the proposed ELM-based method has less training parameters, storage space, offline training and online running time than those of DL-based superimposed CSI feedback.

A Novel Kernel-based Extreme Learning Machine with Incremental Hidden Layer Nodes

Extreme learning machine (ELM) is widely used in various fields because of its advantages such as short training time and good generalization performance. The input weights and bias of hidden layer of traditional ELM are generated randomly, and the number of hidden layer nodes is determined by artificial experience. Only by adjusting parameters manually can an appropriate network structure be found. This training method is complex and time-consuming, which increases the workload of workers. To solve this problem, the incremental extreme learning machine (I-ELM) is used to determine the appropriate number of hidden layer nodes and construct a compact network structure in this paper. At the same time, a new hidden layer activation function STR is proposed, which avoids the disadvantages of incomplete output information of hidden layer due to uneven distribution of sample data. The proposed algorithm is evaluated by public data sets and applied to the classification of superheat degree (SD) in aluminum electrolysis industry. The experimental results show that STR activation function has a good learning speed, and the proposed algorithm is superior to the existing SD identification algorithm in terms of accuracy and robustness.

A Framework for Analyzing, Designing, and Visualizing Spiking Neural Networks—Part II: Nonlinear Response Surfaces

In this second part of a two-part study, we extend to nonlinear synaptic responses a new framework, called Response Surfaces (RSs), for analyzing, designing, and visualizing spiking neurons and networks. An RS is the transfer function between input and output firing times of a spiking neuron and shows all the patterns of input spike times that fire a neuron at a given time. Here in Part II, we present four mathematically tractable functions to model more realistic post-synaptic-potential waveforms, and we build on the linear RS framework of Part I to create a nonlinear RS framework. We then discuss the qualitative differences between the linear and nonlinear RSs frameworks and revisit problems of Part I using the nonlinear RSs framework: graphing the transfer function of a nonlinear spiking neuron, designing an efficient spiking-XOR gate, analyzing phase-tracking in recurrent SNNs, and developing transfer functions for calculating the output firing times of a spiking neuron given a set of input firing times. For the nonlinear RS framework, we show that the output firing time of a nonlinear spiking neuron is equivalent to the center-of-mass of input spike times (acting as positions) and weights (acting as masses) plus a fixed delay plus a second-order correction. The second-order centroid correction is one of the main differences between the linear and nonlinear RS frameworks. For recurrent networks, the nonlinear RS framework also shows a more stable behavior compared to linear SNNs, owing to the smooth shapes of nonlinear post-synaptic-potential waveforms.

An Event-Based Neural Network Architecture with Content Addressable Memory

A hybrid analog/digital very large-scale integration (VLSI) implementation of a spiking neural network with programmable synaptic weights was designed. The synaptic weight values are stored in an asynchronous module, which is interfaced to a fast current-mode event-driven DAC for producing synaptic currents with the appropriate amplitude values. It acts as a transceiver, receiving asynchronous events for input, performing neural computations with hybrid analog/digital circuits on the input spikes, and eventually producing digital asynchronous events in output. Input, output, and synaptic weight values are transmitted to/from the chip using a common communication protocol based on the address event representation (AER). Using this representation, it is possible to interface the device to a workstation or a microcontroller and explore the effect of different types of spike-timing dependent plasticity (STDP) learning algorithms for updating the synaptic weights values in the CAM module.

Extreme Learning Machine-Based Ensemble Transfer Learning for Hyperspectral Image Classification

Although the extreme learning machine (ELM) has been successfully applied to hyperspectral image (HSI) classification, the development of the ELM is restricted by insufficient training data. In this article, we propose a novel extreme learning machine-based ensemble transfer learning algorithm for hyperspectral image classification named TL-ELM. TL-ELM not only retains the input weights and hidden biases of the ELM learned from the target domain, but also utilizes instances in the source domain to iteratively adjust the output weights of the ELM, which are used as the weights of the training models, and then ensembles the training models with their weights for the final classification. In experiments, we choose different regions in northern Italy, namely, Pavia University and Pavia Centre, as the source dataset and target dataset, respectively, and through a comparison with other transfer learning algorithms, we demonstrate that our proposed TL-ELM algorithm is superior on HSI classification tasks with only a few labeled data points in the target domain. Furthermore, we set Pavia University as the source dataset and Pavia Centre as the target dataset to demonstrate that our proposed method can effectively transfer useful instances between different HSIs.

MPC closed-loop identification without excitation

This paper presents a method of closed-loop identification for multivariable systems without external excitation. The method is specially designed for model predictive control (MPC) systems. Without using external excitation (test signals), the method ensures the informativity of the closed-loop data and, at the same time, improve the control performance during the test period. The purpose of the study is to reduce the cost of identification test. The basic idea is to switch the input weighting matrix in the MPC controller which leads to the informativity of the data-set. A preliminary test is carried out in order to find a new input weighting matrix which improve the control performance; then a switching scheme is developed based on the two weighting matrixes. Traditional simulation based model validation no longer works in closed-loop identification without excitation, and model error bounds on the frequency responses can be used instead. The effectiveness of the proposed method is demonstrated by a simulation study.

An Extreme Learning Machine Model Approach on Airbnb Base Price Prediction

The base price of Airbnb properties prediction is still a new area of prediction research, especially with the Extreme Learning Machine (ELM). The previous studies had several suggestions for the advantages of ELM, such as good generalization performance, fast learning speed, and high prediction accuracy. This paper proposes how the ELM approach is used as a prediction model for Air BnB base price. Generally, the steps are setting hidden neuron numbers, randomly assigning input weight and hidden layer biases, calculating the output layer; and the entire learning measure finished through one numerical change without iteration. The performance of the model is estimated utilizing mean squared error, mean absolute percentage error, and root mean squared error. Experiment with Airbnb dataset in London with twenty-one features as input generates a faster learning speed and better accuracy than the existing model.

Toward deep neural networks: Mirror extreme learning machines for pattern classification

In this paper, a novel type of feed-forward neural network with a simple structure is proposed and investigated for pattern classification. Because the novel type of forward neural network?s parameter setting is mirrored with those of the Extreme Learning Machine (ELM), it is termed the mirror extreme learning machine (MELM). For the MELM, the input weights are determined by the pseudoinverse method analytically, while the output weights are generated randomly, which are completely different from the conventional ELM. Besides, a growing method is adopted to obtain the optimal hidden-layer structure. Finally, to evaluate the performance of the proposed MELM, abundant comparative experiments based on different real-world classification datasets are performed. Experimental results validate the high classification accuracy and good generalization performance of the proposed neural network with a simple structure in pattern classification.

Ultra-Short-Term Wind Power Prediction by Salp Swarm Algorithm-Based Optimizing Extreme Learning Machine

Wind power generation accounts for an increasing proportion of the power grid, so efficient and accurate real-time wind power prediction is particularly important for wind power grid. In view of the strong randomness and fluctuation of wind and the difficulty of predicting wind power, a Salp Swarm Algorithms-Extremely Learning Machine (SSA-ELM) based ultra-short-term wind power prediction model is proposed. In this case, the multi-input sample set is composed of historical wind speed, temperature, wind direction, atmospheric pressure and other climatic factors that are highly correlated with wind power, and the network parameters are determined in the training process. In order to improve the adaptability and accuracy of the prediction model, the input weight matrix and hidden layer deviation of the Extreme Learning Machine (ELM) are optimized by exploring and developing the Salp Swarm Algorithm in the iterative process. Finally, the simulation experiment is conducted with the actual data of a wind farm in Henan Province, and the comparison with the traditional Extreme Learning Machine, Particle Swarm Optimization Extreme Learning Machine (PSO-ELM) and Back Propagation (BP) neural network model shows that the new method avoids falling into the local extreme value, and has faster convergence speed and higher prediction accuracy.

Hybrid Convolutional Neural Network with PSO Based Severe Dengue Prognosis Method in Human Genome Data

Dengue is one of the most significant diseases transmitted by arthropods in the world. Dengue phenotypes are focused on documented inaccuracies in the laboratory and clinical studies. In countries with a high incidence of this disease, early diagnosis of dengue is still a concern for public health. Deep learning has been developed as a highly versatile and accurate methodology for classification and regression, which requires small adjustment, interpretable results, and the prediction of risk for complex diseases. This work is motivated by the inclusion of the Particle Swarm Optimization (PSO) algorithm for the fine-tuning of the model's parameters in the convolutional neural network (CNN). The use of this PSO was used to forecast patients with extreme dengue, and to refine the input weight vector and CNN parameters to achieve anticipated precision, and to prevent premature convergence towards local optimum conditions.

Exploring the water quality driving mechanism in Poyang Lake, the largest freshwater lake in China

Abstract Poyang Lake, the largest freshwater lake and typical river-connected lake in China, was selected as a research area. A method was first proposed to quantitatively explore the mechanisms driving water quality evolution, in which the weights of horizontal boundary input, self-purification, vertical atmospheric deposition and sediment release could be determined. A two-dimensional water environment model for Poyang Lake was developed in the framework of the Finite Volume Method and calibrated against the field investigated data. Four typical months in a common-water year were determined for numerical experiments to investigate the temporal and spatial water quality driving mechanisms in Poyang Lake. The results suggested that boundary input and self-purification have the greatest effect on dominating the water quality in Poyang Lake, followed by atmospheric deposition and sediment release. The driving weights of these four factors are 57.2%, 26.5%, 9.3%, and 7.0%, respectively. However, the impact on lake water quality of external water quality, aquatic ecosystem structure, precipitation distribution, and meteorological conditions, which are attributed to the disparities in geographical situation, varied significantly with seasons and locations.

Macroscopic–Microscopic Attention in LSTM Networks Based on Fusion Features for Gear Remaining Life Prediction

In the mechanical transmission system, the gear is one of the most widely used transmission components. The failure of the gear will cause serious accidents and huge economic loss. Therefore, the remaining life prediction of the gear is of great importance. In order to accurately predict the remaining life of the gear, a new type of long-short-term memory neural network with macroscopic–microscopic attention (MMA) is proposed in this article. First, some typical time-domain and frequency-domain characteristics of vibration signals are calculated, respectively, such as the maximum value, the absolute mean value, the standard deviation, the kurtosis, and so on. Then, the principal component of these characteristics is extracted by the isometric mapping method. The importance of fusional characteristic information is filtered via a proposed MMA mechanism so that the input weight of neural network data and recursive data can reach multilevel real-time amplification. With the new long short-term memory neural network, the health characteristics of gear vibration signals can be predicted based on the known fusion features. The experimental results show that this method can predict the remaining life of gears and bearings well, and it has higher prediction accuracy than the conventional prediction methods.

A Cross-Efficiency Evaluation Method Based on Evaluation Criteria Balanced on Interval Weights

Cross-efficiency evaluation approaches and common set of weights (CSW) approaches have long been suggested as two of the more important and effective methods for the ranking of decision making units (DMUs) in data envelopment analysis (DEA). The former emphasizes the flexibility of evaluation and its weights are asymmetric, while the latter focuses on the standardization of evaluation and its weights are symmetrical. As a compromise between these two approaches, this paper proposes a cross-efficiency evaluation method that is based on two types of flexible evaluation criteria balanced on interval weights. The evaluation criteria can be regarded as macro policy—or means of regulation—according to the industry’s current situation. Unlike current cross-efficiency evaluation methods, which tend to choose the set of weights for peer evaluation based on certain preferences, the cross-efficiency evaluation method based on evaluation criterion determines one set of input and output weights for each DMU. This is done by minimizing the difference between the weights of the DMU and the evaluation criteria, thus ensuring that the cross-evaluation of all DMUs for evaluating peers is as consistent as possible. This method also eliminates prejudice and arbitrariness from peer evaluations. As a result, the proposed cross-efficiency evaluation method not only looks for non-zero weights, but also ranks efficient DMUs completely. The proposed DEA model can be further extended to seek a common set of weights for all DMUs. Numerical examples are provided to illustrate the applications of the cross-efficiency evaluation method based on evaluation criterion in DEA ranking.

An Optimal Weight Semi-Supervised Learning Machine for Neural Networks with Time Delay

In this paper, an optimal weight semi-supervised learning machine for a single-hidden layer feedforward network (SLFN) with time delay is developed. Both input weights and output weights of the SLFN are globally optimized with manifold regularization. By feature mapping, input vectors can be placed at the prescribed positions in the feature space in the sense that the separability of all nonlinearly separable patterns can be maximized, unlabeled data can be leveraged to improve the classification accuracy when labeled data are scarce, and a high degree of recognition accuracy can be achieved with a small number of hidden nodes in the SLFN. Some simulation examples are presented to show the excellent performance of the proposed algorithm.

Affective Video Content Analysis With Adaptive Fusion Recurrent Network

Affective video content analysis is an important research topic in video content analysis and has extensive applications. Intuitively, multimodal features can depict elicited emotions, and the accumulation of temporal inputs influences the viewer's emotion. Although a number of research works have been proposed for this task, the adaptive weights of modalities and the correlation of temporal inputs are still not well studied. To address these issues, a novel framework is designed to learn the weights of modalities and temporal inputs from video data. Specifically, three network layers are designed, including statistical-data layer to improve the robustness of data, temporal-adaptive-fusion layer to fuse temporal inputs, and multimodal-adaptive-fusion layer to combine multiple modalities. In particular, the feature vectors of three input modalities are respectively extracted from three pre-trained convolutional neural networks and then fed to three statistical-data layers. Then, the output vectors of these three statistical-data layers are separately connected to three recurrent layers, and the corresponding outputs are fed to a fully-connected layer which shares parameters across modalities and temporal inputs. Finally, the outputs of the fully-connected layer are fused by the temporal-adaptive-fusion layer and then combined by the multimodal-adaptive-fusion layer. To discover the correlation of both multiple modalities and temporal inputs, adaptive weights of modalities and temporal inputs are introduced into loss functions for model training, and these weights are learned by an optimization algorithm. Extensive experiments are conducted on two challenging datasets, which demonstrate that the proposed method achieves better performances than baseline and other state-of-the-art methods.

Slope Stability of Landfill with Waste Degradation

Nowadays, a large amount of municipal solid waste (MSW) is generated due to the rapid urbanisation in developing countries leads to the demand for larger and higher capacity landfills. Bioreactor landfill technology has been introduced to accelerate the stability of landfill and to solve the issue of limited landfill area. However, the accelerated degradation of the refuse in bioreactor landfills also considerably changes the geotechnical characteristics of the waste in the landfill and thereby increases the concern for waste stability. Hence, this study aims to analyse the stability of both conventional and bioreactor landfill slope with the effects of waste degradation. Finite element method has been used in the slope stability analysis and the stability is presented by the factor of safety. The objectives of this study are i) to determine and assess the main parameter which influences the stability of the waste slope, ii) to determine the effects of waste degradation to the waste properties and iii) to obtain the factor of safety of the landfill slope using numerical analysis by finite element method. From the literature review, it is found that slope stability of a landfill mainly depends on the geotechnical properties of waste, such as moisture content, unit weight, shear strength parameters and hydraulic conductivity of waste. After the degradation process, engineering properties of field refuse are affected which includes the increased pore-water pressure and unit weight, decreased strength and lower hydraulic conductivity. Based on the analysis of conventional landfill slope stability by using Plaxis software, slope ratio of 1:3, 1:4 and 1:5 calculated safe with 1.69, 2.3 and 2.8 whereas the analysis of bioreactor landfill slope stability calculated safe only for slope ratio of 1:4 and 1:5 with 1.60 and 1.97. Moreover, the factor of safety for steeper slopes is lower and vice versa. From the parametric analysis, it is found that the full height of slope and unit weight of waste input affect the result analysis. This study is significant to evaluate the landfill slope stability with the effects of waste degradation and to ensure both conventional and bioreactor landfill slope stability for long periods.