As a novel neural network model, the broad learning system (BLS) has been used extensively due to its simple structure and few model hyperparameters. This paper presents a distributed BLS algorithm with quantized and ...
详细信息
Extreme learning machines (ELM) are popular in the field of pattern recognition and machine learning. The kernel extension of ELM (KELM) presents a better performance than traditional ELM. Although the KELM is able to...
详细信息
As a class of recurrent neural networks (RNNs), echo state networks (ESNs) have been studied extensively in recent years, particularly in time-series prediction and non-linear system identification. It is well-known t...
详细信息
Differential evolution algorithm based on the covariance matrix learning can adjust the coordinate system according to the characteristics of the population, which makes the search move in a more favorable direction. ...
详细信息
Differential evolution algorithm based on the covariance matrix learning can adjust the coordinate system according to the characteristics of the population, which makes the search move in a more favorable direction. In order to obtain more accurate information about the function shape, this paper proposescovariance matrix learning differential evolution algorithm based on correlation (denoted as RCLDE)to improve the search efficiency of the algorithm. First, a hybrid mutation strategy is designed to balance the diversity and convergence of the population;secondly, the covariance learning matrix is constructed by selecting the individual with the less correlation;then, a comprehensive learning mechanism is comprehensively designed by two covariance matrix learning mechanisms based on the principle of probability. Finally,the algorithm is tested on the CEC2005, and the experimental results are compared with other effective differential evolution algorithms. The experimental results show that the algorithm proposed in this paper is an effective algorithm.
In this paper, we introduce the concepts of additive generators and additive generator pair of n-dimensional overlap functions, in order to extend the dimensionality of overlap functions from 2 to n. We mainly discuss...
详细信息
In this paper, we introduce the concepts of additive generators and additive generator pair of n-dimensional overlap functions, in order to extend the dimensionality of overlap functions from 2 to n. We mainly discuss the conditions under which an n-dimensional overlap function can be expressed in terms of its generator pair.
In this paper, we firstly introduce some new results on overlap functions and n-dimensional overlap functions. On the other hand, in a previous study, Gómez et al. presented some open problems. One of these open ...
详细信息
In this paper, we firstly introduce some new results on overlap functions and n-dimensional overlap functions. On the other hand, in a previous study, Gómez et al. presented some open problems. One of these open problems is “to search the construction of n-dimensional overlapping functions based on bi-dimensional overlapping functions”. To answer this open problem, in this paper, we mainly introduce one construction method of n-dimensional overlap functions based on bivariate overlap functions. We mainly use the conjunction operator ∧ to construct n-dimensional overlap functions based on bivariate overlap functions and study their basic properties.
As a novel neural network model, the broad learning system (BLS) has been used extensively due to its simple structure and few model hyperparameters. This paper presents a distributed BLS algorithm with quantized and ...
详细信息
ISBN:
(纸本)9781665426480
As a novel neural network model, the broad learning system (BLS) has been used extensively due to its simple structure and few model hyperparameters. This paper presents a distributed BLS algorithm with quantized and censored communications (DQC-BLS). The DQC-BLS algorithm is based on the alternating direction method of multipliers (ADMM). In order to reduce the high communication cost, the DQC-BLS algorithm improves the communication mechanism between the in-networks agents by introducing quantization and communication-censoring strategies. Quantization reduces the number of bits for each transmission whereas communication censoring reduces the number of communications. The experimental results verify that the DQC-BLS algorithm can reduce communication costs while maintaining similar performance on the test set.
Extreme learning machines (ELM) are popular in the field of pattern recognition and machine learning. The kernel extension of ELM (KELM) presents a better performance than traditional ELM. Although the KELM is able to...
详细信息
ISBN:
(纸本)9781665426480
Extreme learning machines (ELM) are popular in the field of pattern recognition and machine learning. The kernel extension of ELM (KELM) presents a better performance than traditional ELM. Although the KELM is able to solve complex nonlinear problems, it is time-consuming and memorydemanding when dealing with a large-size kernel matrix. Introducing reduced kernel technique can dramatically cut down the computational load and memory usage. However, as the amount of training data grows exponentially, it is not effective for a single worker to store the kernel matrix, making it infeasible for data mining in a centralized manner. In this paper, a distributed reduced kernel method for training ELM over decentralized data (DRKELM) is proposed. In the DRKELM, we randomly assign data to different nodes. The communication between nodes is fixed and does not depend on the size of the training data on each node, but on the network topology. Different from the existing reduced kernel ELM, the DRKELM is a fully distributed training algorithm based on the method of alternating direction method of multiplier (ADMM). Experiment with the large scale data set finds that the distributed method can achieve almost the same results as the centralized algorithm and even takes less time to a large extent. It greatly reduces the computation time consumption.
As a class of recurrent neural networks (RNNs), echo state networks (ESNs) have been studied extensively in recent years, particularly in time-series prediction and non-linear system identification. It is well-known t...
详细信息
ISBN:
(纸本)9781665426480
As a class of recurrent neural networks (RNNs), echo state networks (ESNs) have been studied extensively in recent years, particularly in time-series prediction and non-linear system identification. It is well-known that deep learning (DL) has been applied to ESNs and shows more outstanding performance than conventional ESNs. However, due to the large-scale dataset and complexity required to train deep learning models, the time required to train deep ESNs based on traditional centralized algorithms is highly extended. Hence, a decentralized training algorithm combining DL and ESNs is presented in this paper to aim at multiple time scales processing of temporal data. We propose this algorithm based on an optimization program called the alternating direction method of multipliers (ADMM) and decentralized average consensus (DAC) procedure. The results based on a large-scale artificial dataset prove that it has more advantages in generalization accuracy and training time than the centralized experimental scheme.
Non-invasive techniques for rapid blood testing are gaining traction in global healthcare as they optimize medical screening, diagnosis and clinical decisions. Fourier transform infrared (FT-IR) spectroscopy is one of...
详细信息
Non-invasive techniques for rapid blood testing are gaining traction in global healthcare as they optimize medical screening, diagnosis and clinical decisions. Fourier transform infrared (FT-IR) spectroscopy is one of the most common technologies that can be used for non-destructive aided medical detection. Typically, after acquiring the Fourier transform infrared spectrum, spectral data preprocessing and feature extraction and quantitative analysis of several indicators of blood samples can be accomplished, in combination with chemometric method studies. At present, blood hemoglobin (HGB) concentration is one of the most valuable information for the clinical diagnosis of patient’s health status. FT-IR spectroscopy is employed as a green technique aided medical test of blood HGB. Then the acquired HGB concentration data is switched to the spectral feature data by the studies of advanced chemometric method, in help for hiding the sensitive medical information to protect the privacy of patients. The decision tree network architecture is proposed for feature extraction of FT-IR data in order to find the small set of wavenumbers that are able to quantify HGB. A semi-supervised learning strategy is designed for tuning the number of network neuron nodes, in the way of searching for the maximum entropy increment. Each neuron is optimized by the growing of a semi-supervised decision tree, to accurately identify the informative FT-IR wavenumbers. The features extracted by the semi-supervised learning decision tree network guarantees the FT-IR aided detection model has high efficiency and high prediction accuracy. A model of quantifying the HGB concentration shows that the proposed decision tree network with semi-supervised entropy learning strategy outperforms the usual methods of full spectrum partial least square model and the fully connected neural network model in prediction accuracy. The framework is expected to support the FT-IR spectral technology for aided detection
暂无评论