Traditional Learning Algorithms Research Articles

LLE-based classification algorithm for MMW radar target recognition

In this paper, a new multiclass classification algorithm is proposed based on the idea of Locally Linear Embedding (LLE), to avoid the defect of traditional manifold learning algorithms, which can not deal with new sample points. The algorithm defines an error as a criterion by computing a sample’s reconstruction weight using LLE. Furthermore, the existence and characteristics of low dimensional manifold in range-profile time-frequency information are explored using manifold learning algorithm, aiming at the problem of target recognition about high range resolution MilliMeter-Wave (MMW) radar. The new algorithm is applied to radar target recognition. The experiment results show the algorithm is efficient. Compared with other classification algorithms, our method improves the recognition precision and the result is not sensitive to input parameters.

Fault diagnosis for ship sewage treatment equipment based on clustering support vector machine

To solve the problems of condition monitoring and fault diagnosis for ship sewage treatment equipment, a fault diagnosis model based on clustering Support Vector Machines (SVM) was proposed. In the model, the neural network clustering algorithm was used to realize clustering analysis and to obtain the normal subspaces and the abnormal subspaces in the condition monitoring sample space. Then, to the abnormal subspaces, the multi-classification SVM based on binary tree architecture was designed to carry out the fault diagnosis and recognition. Compared with the traditional SVM learning algorithms, the proposed algorithm avoided the blind classification and improved the classification performance in some extent. The model was applied to a ship sewage treatment equipment to train and exam the measured samples. Experiment results show this method has good generalization and expendibility.

Boosting feature selection using information metric for classification

Feature selection plays an important role in pattern classification. Its purpose is to remove redundant features from data set as many as possible. The presence of useless features may not only deteriorate the performance of learning algorithms, but also obscure important information (e.g., intrinsic structure) behind data. Along with new and emerging techniques, data sets in many domains are becoming larger and larger and many irrelevant features are often prevailing in these data sets. This, however, poses great challenges to traditional learning algorithms, such as low efficiency and over-fitting. Thus, it becomes apparent that an efficient technique is needed to eliminate redundant or irrelevant features from the data sets. Currently, many endeavors to cope with this problem have been attempted and various outstanding feature selection methods have been proposed. Unlike other selection methods, in this paper we propose a general scheme of boosting feature selection method using information metric. The primary characteristic of our method is that it exploits weight of data to select salient features. Furthermore, the weight of data will be dynamically changed after each candidate feature has been selected. Thus, the information criteria used in feature selector can exactly represent the relevant degree between features and the class labels. As a result, the selected feature subset has maximal relevance to the class labels. Simulation studies carried out on UCI data sets show that the classification performance achieved by our proposed method is better than those of other selection methods in most cases.

The attention-gated reinforcement learning model can explain experimentally observed changes in tuning curves that follow category learning

A problem with many traditional neural network learning algorithms is that they lack a biologically plausible method of assigning credit to the nodes in the earlier levels of the network that played a decisive role in the stimulus response mapping. In the Attention-Gated Reinforcement Learning Model (AGREL) [1] the credit assignment problem is overcome by the focal influence of attention through the feedback connections in the network. In each trial, a global reward value (conveyed by neuromodulators) is calculated and serves to increase the likelihood that successful behavior be repeated. The activity in the output level of the network (which is limited so that only a single node at a time may be active/attended) is back propagated through the network via feedback connections. The feedback connections, which are reciprocal in strength to the feedforward connections, allow the network to selectively target the lower level nodes that drove the current output, which in combination with the global reward factor, allows the network to enhance connections with nodes that caused successful behaviors, see figure 1. This attentional gating of the reward signal offers a biologically realistic solution to the credit assignment problem and in so doing provides a coherent and unifying framework for learning, with attentional selection at its core. AGREL has previously been shown to closely replicate the changes in tuning curves observed in two physiological categorization tasks [1]. In the current study, we first replicate the findings of two further physiological tasks using AGREL, and second, examine the effect of removing the feedback connections, on our models performance in these and the previous tasks.

A Support-Vector-Machine-Based Approach to RF Sensor Spectral Signature Classifications

In this paper, a support vector machine (SVM) classifier was designed to identify tornado vortices based on their characteristics that were determined from the Doppler spectra and eigenvalues that were calculated from the data that were collected in the vicinity of these vortices. To collect these data, weather surveillance radar (WSR-88D) was employed, which is locally operated by the National Severe Storms Laboratory (NSSL) on the north campus of the University of Oklahoma, Norman. This particular radar, which is devoted to experimental research rather than operational weather forecasting, has the unique capability of recording massive volumes of in-phase (I) and quadrature (Q) data over many hours. As such, the received radar echo power and Doppler shift information can be determined from these I and Q complex-valued data, which provides a rich environment for evaluating our new postprocessing algorithms. At the current time, most radar systems, including North America's national network of weather radar systems, process these I and Q data in real time to produce environmental measurements known as ldquospectral moments,rdquo which are the first three statistical moments of the data. These three moments (the reflectivity, the radial velocity, and the spectrum width) are then provided to scientists or forecasters, and the abundance of I and Q data is not preserved to save computer resources. One of the aims of this paper is to show how additional information about the atmosphere can be garnered from the I and Q data. To assist this mission, an SVM-based classifier evades the pitfalls of the traditional statistical learning algorithms, such as neural networks, by setting up a convex optimization problem with a single global minimum. In addition, through the use of kernels and a nonlinear mapping to higher dimensional spaces, the SVM classifier is able to effectively handle nonlinear classification problems. The idea behind this transformation is to facilitate the separability of classes by taking the input vectors to a higher dimensional space. The SVM classifier has the added advantage of reducing overfitting by constructing a maximum margin to separate hyperplanes in a higher dimensional feature space to ensure a small generalization error bound. Finally, our practical results are in positive agreement with our theoretical predictions.

Using promoters and functional introns in genetic algorithms for neuroevolutionary learning in non-stationary problems

This paper addresses the problem of adaptive learning in non-stationary problems through neuroevolution. It is a general problem that is very relevant in many tasks, for example, in the context of robot model learning from interaction with the world. Traditional learning algorithms fail in this task as they have mostly been designed for learning a single model in a static setting. Neuroevolutionary techniques have obtained promising results in this non-stationary context but are still lacking in certain types of problems, especially those dealing with information streams where different portions correspond to different models. An extension through the introduction of the concept of introns and promoter genes enables neuroevolutionary algorithms to improve their performance on this type of problems. Following this approach, an implementation of these concepts on a genetic algorithm for neuroevolution is presented here. This algorithm is called promoter based genetic algorithm (PBGA) and it uses a genotypic representation with a set of features that allows for an intrinsic memory in the population that is self-regulated, in the sense that functional parts of the individuals are preserved through generations without an explicit knowledge about the number of different tasks or models that have to arise from the data stream. Some illustrative tests of the potential of these techniques based on the continuous switch between completely different objective functions that must be learnt are presented and the results are analyzed and compared to other neuroevolutionary algorithms.

Framework and implications of virtual neurorobotics

Despite decades of societal investment in artificial learning systems, truly “intelligent” systems have yet to be realized. These traditional models are based on input-output pattern optimization and/or cognitive production rule modeling. One response has been social robotics, using the interaction of human and robot to capture important cognitive dynamics such as cooperation and emotion; to date, these systems still incorporate traditional learning algorithms. More recently, investigators are focusing on the core assumptions of the brain “algorithm” itself—trying to replicate uniquely “neuromorphic” dynamics such as action potential spiking and synaptic learning. Only now are large-scale neuromorphic models becoming feasible, due to the availability of powerful supercomputers and an expanding supply of parameters derived from research into the brain's interdependent electrophysiological, metabolomic and genomic networks. Personal computer technology has also led to the acceptance of computer-generated humanoid images, or “avatars”, to represent intelligent actors in virtual realities. In a recent paper, we proposed a method of virtual neurorobotics (VNR) in which the approaches above (social-emotional robotics, neuromorphic brain architectures, and virtual reality projection) are hybridized to rapidly forward-engineer and develop increasingly complex, intrinsically intelligent systems. In this paper, we synthesize our research and related work in the field and provide a framework for VNR, with wider implications for research and practical applications.

Efficient Computation and Model Selection in Semi-Supervised Learning

Traditional learning algorithm uses only labeled data for training. However, labeled examples are often difficult or time consuming to obtain since they require substantial labeling efforts from humans. On the other hand, unlabeled data are often relatively easy to collect. Semi-supervised learning addresses this problem by using large quantities of unlabeled data with the labeled data to build better learning algorithms. In this paper, we propose a general approach augmenting traditional supervised learning into semi-supervised learning paradigm. A regularization framework which balances a tradeoff between loss and penalty is established. We investigate different implementations of loss function and suggest the methods which have the least computation expenses. The value of a hyperparameter, which determines the balance between loss and penalty, is crucial in model selection. Hence, we derive an algorithm that can fit the entire path of solutions for every value of hyperparameter. Its computational complexity is quadratic in the number of labeled examples only rather than the total number of labeled and unlabeled examples.

Fast Conditional Independence-based Bayesian Classifier

Machine Learning (ML) has become very popular within Data Mining (KDD) and Artificial Intelligence (AI) research and their applications. In the ML and KDD contexts, two main approaches can be used for inducing a Bayesian Network (BN) from data, namely, Conditional Independence (CI) and the Heuristic Search (HS). When a BN is induced for classification purposes (Bayesian Classifier - BC), it is possible to impose some specific constraints aiming at increasing the computational efficiency. In this paper a new CI based approach to induce BCs from data is proposed and two algorithms are presented. Such approach is based on the Markov Blanket concept in order to impose some constraints and optimize the traditional PC learning algorithm. Experiments performed with the ALARM, as well as other six UCI and three artificial domains revealed that the proposed approach tends to execute fewer comparison tests than the traditional PC. The experiments also show that the proposed algorithms produce competitive classification rates when compared with both, PC and Naive Bayes.

Kernel quadratic discriminant analysis for small sample size problem

It is generally believed that quadratic discriminant analysis (QDA) can better fit the data in practical pattern recognition applications compared to linear discriminant analysis (LDA) method. This is due to the fact that QDA relaxes the assumption made by LDA-based methods that the covariance matrix for each class is identical. However, it still assumes that the class conditional distribution is Gaussian which is usually not the case in many real-world applications. In this paper, a novel kernel-based QDA method is proposed to further relax the Gaussian assumption by using the kernel machine technique. The proposed method solves the complex pattern recognition problem by combining the QDA solution and the kernel machine technique, and at the same time, tackles the so-called small sample size problem through a regularized estimation of the covariance matrix. Extensive experimental results indicate that the proposed method is a more sophisticated solution outperforming many traditional kernel-based learning algorithms.

Case-based myopic reinforcement learning for satisfying target service level in supply chain

In the last decade, driven by global competition in the marketplace, many companies have taken initiatives to revamp their supply chains in order to increase responsiveness to changes in the marketplace. The renovation of inventory control system is central to such an effort. However, experiences in industry have shown that the control of inventory in supply chain is not an easy task because of uncertainties inherent in customer demand. In this paper, we propose a reinforcement learning algorithm appropriate for the nonstationary inventory control problem of supply chain that has a large state space. Traditional reinforcement learning algorithms such as learning automata and Q-learning have the difficulty of slow convergence when applied to the situations with large state spaces. To resolve the problems of nonstationary customer demand and large state space, we develop a case-based myopic reinforcement learning (CMRL) algorithm. A simulation-based experiment was performed to show good performance of CMRL.

Multi-stage extreme learning machine for fault diagnosis on hydraulic tube tester

The running status of hydraulic tube tester is reflected by the boosting pressure curve in Hydrostatic testing process. The authors present the extreme learning machine (ELM), a novel good learning scheme much faster than traditional gradient-based learning algorithms, as a mechanism for clustering the pressure curves. However, it caused low accuracy for clustering pressure curves for hydraulic tube tester. In this paper, a multi-stage ELM is proposed to improve the accuracy of clustering. During the process of this new ELM, the input data were divided into several stages, then, every stage was analyzed independently. At last, this method has been used in hydraulic tube tester data. Compared with individual ELM, it has better function for considering the characteristics of input data.

Nonlinear Discriminant Analysis on Embedded Manifold

Traditional manifold learning algorithms, such as ISOMAP, LLE, and Laplacian Eigenmap, mainly focus on uncovering the latent low-dimensional geometry structure of the training samples in an unsupervised manner where useful class information is ignored. Therefore, the derived low-dimensional representations are not necessarily optimal in discriminative capability. In this paper, we study the discriminant analysis problem by considering the nonlinear manifold structure of data space. To this end, firstly, a new clustering algorithm, called Intra-Cluster Balanced K-Means (ICBKM), is proposed to partition the samples into multiple clusters while ensure that there are balanced samples for the classes within each cluster; approximately, each cluster can be considered as a local patch on the embedded manifold. Then, the local discriminative projections for different clusters are simultaneously calculated by optimizing the global Fisher Criterion based on the cluster weighted data representation. Compared with traditional linear/kernel discriminant analysis (KDA) algorithms, our proposed algorithm has the following characteristics: 1) it essentially is a KDA algorithm with specific geometry-adaptive-kernel tailored to the specific data structure, in contrast to traditional KDA in which the kernel is fixed and independent to the data set; 2) it is approximately a locally linear while globally nonlinear discriminant analyzer; 3) it does not need to store the original samples for computing the low-dimensional representation of a new data; and 4) it is computationally efficient compared with traditional KDA when the sample number is large. The toy problem on artificial data demonstrates the effectiveness of our proposed algorithm in deriving discriminative representations for problems with nonlinear classification hyperplane. The face recognition experiments on YALE and CMU PIE databases show that our proposed algorithm significantly outperforms linear discriminant analysis (LDA) as well as Mixture LDA, and has higher accuracy than KDA with traditional kernels

Multilinear Discriminant Analysis for Face Recognition

There is a growing interest in subspace learning techniques for face recognition; however, the excessive dimension of the data space often brings the algorithms into the curse of dimensionality dilemma. In this paper, we present a novel approach to solve the supervised dimensionality reduction problem by encoding an image object as a general tensor of second or even higher order. First, we propose a discriminant tensor criterion, whereby multiple interrelated lower dimensional discriminative subspaces are derived for feature extraction. Then, a novel approach, called k-mode optimization, is presented to iteratively learn these subspaces by unfolding the tensor along different tensor directions. We call this algorithm multilinear discriminant analysis (MDA), which has the following characteristics: 1) multiple interrelated subspaces can collaborate to discriminate different classes, 2) for classification problems involving higher order tensors, the MDA algorithm can avoid the curse of dimensionality dilemma and alleviate the small sample size problem, and 3) the computational cost in the learning stage is reduced to a large extent owing to the reduced data dimensions in k-mode optimization. We provide extensive experiments on ORL, CMU PIE, and FERET databases by encoding face images as second- or third-order tensors to demonstrate that the proposed MDA algorithm based on higher order tensors has the potential to outperform the traditional vector-based subspace learning algorithms, especially in the cases with small sample sizes.

SOIL MOISTURE PREDICTION USING SUPPORT VECTOR MACHINES

ABSTRACT: Herein, a recently developed methodology, Support Vector Machines (SVMs), is presented and applied to the challenge of soil moisture prediction. Support Vector Machines are derived from statistical learning theory and can be used to predict a quantity forward in time based on training that uses past data, hence providing a statistically sound approach to solving inverse problems. The principal strength of SVMs lies in the fact that they employ Structural Risk Minimization (SRM) instead of Empirical Risk Minimization (ERM). The SVMs formulate a quadratic optimization problem that ensures a global optimum, which makes them superior to traditional learning algorithms such as Artificial Neural Networks (ANNs). The resulting model is sparse and not characterized by the “curse of dimensionality.” Soil moisture distribution and variation is helpful in predicting and understanding various hydrologic processes, including weather changes, energy and moisture fluxes, drought, irrigation scheduling, and rainfall/runoff generation. Soil moisture and meteorological data are used to generate SVM predictions for four and seven days ahead. Predictions show good agreement with actual soil moisture measurements. Results from the SVM modeling are compared with predictions obtained from ANN models and show that SVM models performed better for soil moisture forecasting than ANN models.

CPBUM neural networks for modeling with outliers and noise

In this study, CPBUM neural networks with annealing robust learning algorithm (ARLA) are proposed to improve the problems of conventional neural networks for modeling with outliers and noise. In general, the obtained training data in the real applications maybe contain the outliers and noise. Although the CPBUM neural networks have fast convergent speed, these are difficult to deal with outliers and noise. Hence, the robust property must be enhanced for the CPBUM neural networks. Additionally, the ARLA can be overcome the problems of initialization and cut-off points in the traditional robust learning algorithm and deal with the model with outliers and noise. In this study, the ARLA is used as the learning algorithm to adjust the weights of the CPBUM neural networks. It tunes out that the CPBUM neural networks with the ARLA have fast convergent speed and robust against outliers and noise than the conventional neural networks with robust mechanism. Simulation results are provided to show the validity and applicability of the proposed neural networks.

Multi-robot concurrent learning of cooperative behaviours for the tracking of multiple moving targets

Reinforcement learning has been extensively studied and applied for generating cooperative behaviours in multi-robot systems. However, traditional reinforcement learning algorithms assume discrete state and action spaces with finite number of elements. This limits the learning to discrete behaviours and cannot be applied to most real multi-robot systems that inherently require appropriate combinations of different elementary behaviours. To address this problem, we design a distributed learning controller that integrates reinforcement learning with behaviour-based control networks. This learning controller can enable the robots to generate appropriate control policy without the need for human design or hardcoding. Furthermore, to address the problems in concurrent learning, we propose a distributed learning control algorithm to coordinate the concurrent learning processes. The distributed learning controller and learning control algorithm are applied to multi-robot tracking of multiple moving targets. The efficacy of our proposed scheme is shown through simulations.

Learning probabilistic decision trees for AUC

Accurate ranking, measured by AUC (the area under the ROC curve), is crucial in many real-world applications. Most traditional learning algorithms, however, aim only at high classification accuracy. It has been observed that traditional decision trees produce good classification accuracy but poor probability estimates. Since the ranking generated by a decision tree is based on the class probabilities, a probability estimation tree (PET) with accurate probability estimates is desired in order to yield high AUC. Some researchers ascribe the poor probability estimates of decision trees to the decision tree learning algorithms. To our observation, however, the representation also plays an important role. In this paper, we propose to extend decision trees to represent a joint distribution and conditional independence, called conditional independence trees (CITrees), which is a more suitable model for yielding high AUC. We propose a novel AUC-based algorithm for learning CITrees, and our experiments show that the CITree algorithm outperforms the state-of-the-art decision tree learning algorithm C4.4 (a variant of C4.5), naive Bayes, and NBTree in AUC. Our work provides an effective model and algorithm for applications in which an accurate ranking is required.

A Multiple Instance Learning Problem Approach Model to Anomaly Network Intrusion Detection

Even though mainly statistical methods have been used in anomaly network intrusion detection, to detect various attack types, machine learning based anomaly detection was introduced. Machine learning based anomaly detection started from research applying traditional learning algorithms of artificial intelligence to intrusion detection. However, detection rates of these methods are not satisfactory. Especially, high false positive and repeated alarms about the same attack are problems. The main reason for this is that one packet is used as a basic learning unit. Most attacks consist of more than one packet. In addition, an attack does not lead to a consecutive packet stream. Therefore, with grouping of related packets, a new approach of group-based learning and detection is needed. This type of approach is similar to that of multiple-instance problems in the artificial intelligence community, which cannot clearly classify one instance, but classification of a group is possible. We suggest group generation algorithm grouping related packets, and a learning algorithm based on a unit of such group. To verify the usefulness of the suggested algorithm, 1998 DARPA data was used and the results show that our approach is quite useful.

Evolutionary extreme learning machine

Extreme learning machine (ELM) [G.-B. Huang, Q.-Y. Zhu, C.-K. Siew, Extreme learning machine: a new learning scheme of feedforward neural networks, in: Proceedings of the International Joint Conference on Neural Networks (IJCNN2004), Budapest, Hungary, 25–29 July 2004], a novel learning algorithm much faster than the traditional gradient-based learning algorithms, was proposed recently for single-hidden-layer feedforward neural networks (SLFNs). However, ELM may need higher number of hidden neurons due to the random determination of the input weights and hidden biases. In this paper, a hybrid learning algorithm is proposed which uses the differential evolutionary algorithm to select the input weights and Moore–Penrose (MP) generalized inverse to analytically determine the output weights. Experimental results show that this approach is able to achieve good generalization performance with much more compact networks.