An incremental neural learning framework and its application to vehicle diagnostics

This paper presents a framework for incremental neural learning (INL) that allows a base neural learning system to incrementally learn new knowledge from only new data without forgetting the existing knowledge. Upon subsequent encounters of new data examples, INL utilizes prior knowledge to direct its incremental learning. A number of critical issues are addressed including when to make the system learn new knowledge, how to learn new knowledge without forgetting existing knowledge, how to perform inference using both the existing and the newly learnt knowledge, and how to detect and deal with aged learnt systems. To validate the proposed INL framework, we use backpropagation (BP) as a base learner and a multi-layer neural network as a base intelligent system. INL has several advantages over existing incremental algorithms: it can be applied to a broad range of neural network systems beyond the BP trained neural networks; it retains the existing neural network structures and weights even during incremental learning; the neural network committees generated by INL do not interact with one another and each sees the same inputs and error signals at the same time; this limited communication makes the INL architecture attractive for parallel implementation. We have applied INL to two vehicle fault diagnostics problems: end-of-line test in auto assembly plants and onboard vehicle misfire detection. These experimental results demonstrate that the INL framework has the capability to successfully perform incremental learning from unbalanced and noisy data. In order to show the general capabilities of INL, we also applied INL to three general machine learning benchmark data sets. The INL systems showed good generalization capabilities in comparison with other well known machine learning algorithms.     

混合式贝叶斯网络结构增量学习算法研究

提出一种混合式贝叶斯网络结构增量学习算法.首先提出多项式时间的限制性学习技术,为每个变量建立候选父节点集合;然后,依据候选父节点集合,利用搜索技术对当前网络进行增量学习.该算法的复杂度显著低于目前最优的贝叶斯网络增量学习算法.理论与实验均表明,所处理的问题越复杂,该算法在计算复杂度方面的优势越明显.     

Selective negative correlation learning approach to incremental learning

Negative correlation learning (NCL) is a successful approach to constructing neural network ensembles. In batch learning mode, NCL outperforms many other ensemble learning approaches. Recently, NCL has also shown to be a potentially powerful approach to incremental learning, while the advantages of NCL have not yet been fully exploited. In this paper, we propose a selective NCL (SNCL) algorithm for incremental learning. Concretely, every time a new training data set is presented, the previously trained neural network ensemble is cloned. Then the cloned ensemble is trained on the new data set. After that, the new ensemble is combined with the previous ensemble and a selection process is applied to prune the whole ensemble to a fixed size. This paper is an extended version of our preliminary paper on SNCL. Compared to the previous work, this paper presents a deeper investigation into SNCL, considering different objective functions for the selection process and comparing SNCL to other NCL-based incremental learning algorithms on two more real world bioinformatics data sets. Experimental results demonstrate the advantage of SNCL. Further, comparisons between SNCL and other existing incremental learning algorithms, such Learn++ and ARTMAP, are also presented.     

A robust incremental learning framework for accurate skin region segmentation in color images

In this paper, we propose a robust incremental learning framework for accurate skin region segmentation in real-life images. The proposed framework is able to automatically learn the skin color information from each test image in real-time and generate the specific skin model (SSM) for that image. Consequently, the SSM can adapt to a certain image, in which the skin colors may vary from one region to another due to illumination conditions and inherent skin colors. The proposed framework consists of multiple iterations to learn the SSM, and each iteration comprises two major steps: (1) collecting new skin samples by region growing; (2) updating the skin model incrementally with the available skin samples. After the skin model converges (i.e., becomes the SSM), a post-processing can be further performed to fill up the interstices on the skin map. We performed a set of experiments on a large-scale real-life image database and our method observably outperformed the well-known Bayesian histogram. The experimental results confirm that the SSM is more robust than static skin models.     

DetectA: abrupt concept drift detection in non-stationary environments

Almost all drift detection mechanisms designed for classification problems work reactively: after receiving the complete data set (input patterns and class labels) they apply a sequence of procedures to identify some change in the class-conditional distribution – a concept drift. However, detecting changes after its occurrence can be in some situations harmful to the process under analysis. This paper proposes a proactive approach for abrupt drift detection, called DetectA (Detect Abrupt Drift). Briefly, this method is composed of three steps: (i) label the patterns from the test set (an unlabelled data block), using an unsupervised method; (ii) compute some statistics from the train and test sets, conditioned to the given class labels for train set; and (iii) compare the training and testing statistics using a multivariate hypothesis test. Based on the results of the hypothesis tests, we attempt to detect the drift on the test set, before the real labels are obtained. A procedure for creating datasets with abrupt drift has been proposed to perform a sensitivity analysis of the DetectA model. The result of the sensitivity analysis suggests that the detector is efficient and suitable for datasets of high-dimensionality, blocks with any proportion of drifts, and datasets with class imbalance. The performance of the DetectA method, with different configurations, was also evaluated on real and artificial datasets, using an MLP as a classifier. The best results were obtained using one of the detection methods, being the proactive manner a top contender regarding improving the underlying base classifier accuracy.     

A scalable, incremental learning algorithm for classification problems

In this paper a novel data mining algorithm, Clustering and Classification Algorithm-Supervised (CCA-S), is introduced. CCA-S enables the scalable, incremental learning of a non-hierarchical cluster structure from training data. This cluster structure serves as a function to map the attribute values of new data to the target class of these data, that is, classify new data. CCA-S utilizes both the distance and the target class of training data points to derive the cluster structure. In this paper, we first present problems with many existing data mining algorithms for classification problems, such as decision trees, artificial neural networks, in scalable and incremental learning. We then describe CCA-S and discuss its advantages in scalable, incremental learning. The testing results of applying CCA-S to several common data sets for classification problems are presented. The testing results show that the classification performance of CCA-S is comparable to the other data mining algorithms such as decision trees, artificial neural networks and discriminant analysis.     

Superimposing neural learning by dynamic and spatial changing weights

In this article, a new neural network model is presented for incremental learning tasks where networks are required to learn new knowledge without forgetting the old. An essential core of the proposed network structure is their dynamic and spatial changing connection weights (DSCWs). A learning scheme is developed for the formulation of the dynamic changing weights, while a structural adaptation is formulated by the spatial changing connecting weights. To avoid disturbing the old knowledge by the creation of new connections, a restoration mechanism is introduced dusing the DSCWs. The usefulness of the proposed model is demonstrated by using a system identification task. This work was presented in part at the 7th International Symposium on Artificial Life and Robotics, Oita, Japan, January 16–18, 2002.     

Negative correlation in incremental learning

Negative Correlation Learning (NCL) has been successfully applied to construct neural network ensembles. It encourages the neural networks that compose the ensemble to be different from each other and, at the same time, accurate. The difference among the neural networks that compose an ensemble is a desirable feature to perform incremental learning, for some of the neural networks can be able to adapt faster and better to new data than the others. So, NCL is a potentially powerful approach to incremental learning. With this in mind, this paper presents an analysis of NCL, aiming at determining its weak and strong points to incremental learning. The analysis shows that it is possible to use NCL to overcome catastrophic forgetting, an important problem related to incremental learning. However, when catastrophic forgetting is very low, no advantage of using more than one neural network of the ensemble to learn new data is taken and the test error is high. When all the neural networks are used to learn new data, some of them can indeed adapt better than the others, but a higher catastrophic forgetting is obtained. In this way, it is important to find a trade-off between overcoming catastrophic forgetting and using an entire ensemble to learn new data. The NCL results are comparable with other approaches which were specifically designed to incremental learning. Thus, the study presented in this work reveals encouraging results with negative correlation in incremental learning, showing that NCL is a promising approach to incremental learning.

Fast incremental learning methods inspired by biological learning behavior

Model-based learning systems such as neural networks usually “forget” learned skills due to incremental learning of new instances. This is because the modification of a parameter interferes with old memories. Therefore, to avoid forgetting, incremental learning processes in these learning systems must include relearning of old instances. The relearning process, however, is time-consuming. We present two types of incremental learning method designed to achieve quick adaptation with low resources. One approach is to use a sleep phase to provide time for learning. The other one involves a “meta-learning module” that acquires learning skills through experience. The system carries out “reactive modification” of parameters not only to memorize new instances, but also to avoid forgetting old memories using a meta-learning module.This work was presented, in part, at the 9th International Symposium on Artificial Life and Robotics, Oita, Japan, January 28–30, 2004     

A robust and flexible model of hierarchical self-organizing maps for non-stationary environments

In this paper we extend the hierarchical self-organizing maps model (HSOM) to address the problem of learning topological drift under non-stationary and noisy environments. The new model, called robust and flexible hierarchical self-organizing maps (RoFlex-HSOM), combines the capabilities of robustness against noise and the flexibility to adapt to the changing environment.The RoFlex-HSOM model consists of a hierarchical tree structure of growing self-organizing maps (SOMs) that adapts its architecture based on the data. The model preserves the topology mapping from the high-dimensional time-dependent input space onto a neuron position in a low-dimensional hierarchical output space grid. Furthermore, the RoFlex-HSOM algorithm has the plasticity to track and adapt to the topological drift, it gradually forgets (but no catastrophically) previous learned patterns and it is resistant to the presence of noise. We empirically show the capabilities of our model with experimental results using synthetic sequential data sets and the “El Niño” real world data.     

On incremental and robust subspace learning

Principal Component Analysis (PCA) has been of great interest in computer vision and pattern recognition. In particular, incrementally learning a PCA model, which is computationally efficient for large-scale problems as well as adaptable to reflect the variable state of a dynamic system, is an attractive research topic with numerous applications such as adaptive background modelling and active object recognition. In addition, the conventional PCA, in the sense of least mean squared error minimisation, is susceptible to outlying measurements. To address these two important issues, we present a novel algorithm of incremental PCA, and then extend it to robust PCA. Compared with the previous studies on robust PCA, our algorithm is computationally more efficient. We demonstrate the performance of these algorithms with experimental results on dynamic background modelling and multi-view face modelling.     

Info-fuzzy algorithms for mining dynamic data streams

Most data-mining algorithms assume static behavior of the incoming data. In the real world, the situation is different and most continuously collected data streams are generated by dynamic processes, which may change over time, in some cases even drastically. The change in the underlying concept, also known as concept drift, causes the data-mining model generated from past examples to become less accurate and relevant for classifying the current data. Most online learning algorithms deal with concept drift by generating a new model every time a concept drift is detected. On one hand, this solution ensures accurate and relevant models at all times, thus implying an increase in the classification accuracy. On the other hand, this approach suffers from a major drawback, which is the high computational cost of generating new models. The problem is getting worse when a concept drift is detected more frequently and, hence, a compromise in terms of computational effort and accuracy is needed. This work describes a series of incremental algorithms that are shown empirically to produce more accurate classification models than the batch algorithms in the presence of a concept drift while being computationally cheaper than existing incremental methods. The proposed incremental algorithms are based on an advanced decision-tree learning methodology called “Info-Fuzzy Network” (IFN), which is capable to induce compact and accurate classification models. The algorithms are evaluated on real-world streams of traffic and intrusion-detection data.     

Convex incremental extreme learning machine

Unlike the conventional neural network theories and implementations, Huang et al. [Universal approximation using incremental constructive feedforward networks with random hidden nodes, IEEE Transactions on Neural Networks 17(4) (2006) 879–892] have recently proposed a new theory to show that single-hidden-layer feedforward networks (SLFNs) with randomly generated additive or radial basis function (RBF) hidden nodes (according to any continuous sampling distribution) can work as universal approximators and the resulting incremental extreme learning machine (I-ELM) outperforms many popular learning algorithms. I-ELM randomly generates the hidden nodes and analytically calculates the output weights of SLFNs, however, I-ELM does not recalculate the output weights of all the existing nodes when a new node is added. This paper shows that while retaining the same simplicity, the convergence rate of I-ELM can be further improved by recalculating the output weights of the existing nodes based on a convex optimization method when a new hidden node is randomly added. Furthermore, we show that given a type of piecewise continuous computational hidden nodes (possibly not neural alike nodes), if SLFNs can work as universal approximators with adjustable hidden node parameters, from a function approximation point of view the hidden node parameters of such “generalized” SLFNs (including sigmoid networks, RBF networks, trigonometric networks, threshold networks, fuzzy inference systems, fully complex neural networks, high-order networks, ridge polynomial networks, wavelet networks, etc.) can actually be randomly generated according to any continuous sampling distribution. In theory, the parameters of these SLFNs can be analytically determined by ELM instead of being tuned.     

Evaluation of robust control by universal learning networks

The characteristics of control system design using a universal learning network (ULN) are such that both the controlled systems and their controller are represented in a unified framework, and that the learning stage of the ULN can be executed by using not only first-order derivatives (gradient) but also the higher order derivatives of the criterion function with respect to parameters. ULNs have the same generalization ability as neural networks. So the ULN controller is able to control the system in a favorable way under an environment which is little different from the environment of the control system at the learning stage. However, stability cannot be sufficiently realized. In this paper, we propose a robust control method using a ULN and second-order derivatives of that ULN. Robust control, as considered here, is defined as follows. Even though the initial values of the node outputs are very different from those at the learning stage, the control system is able to reduce its influence to other node outputs and can control the system as in the case of no variation. In order to realize such robust control, a new term concerning the variation is added to the usual criterion function, and the parameters are adjusted so as to minimize the above-mentioned criterion function using second-order derivatives of the criterion function with respect to the parameters. Finally, it is shown that the ULN controller constructed by the proposed method works effectively in a simulation study of a non-linear crane system. This work was presented, in part, at the International Symposium on Artificial Life and Robotics, Oita, Japan, February 18–20, 1996     

一种新的类增量学习方法

提出一种新的基于超椭球的类增量学习算法。对每一类样本,在特征空间求得一个包围该类尽可能多样本的最小超椭球,使得各类样本之间通过超椭球隔开。类增量学习过程中,只对新增类样本进行训练。分类时,通过计算待分类样本是否在超椭球内判定其所属类别。实验结果证明,该方法较超球方法提高了分类精度和分类速度。     

A filter based neuron model for adaptive incremental learning of self-organizing maps

In this work a learning algorithm is proposed for the formation of topology preserving maps. In the proposed algorithm the weights are updated incrementally using a higher-order difference equation, which implements a low-pass digital filter. It is shown that by suitably choosing the filter the learning process can adaptively follow a specific dynamic. Numerical results, for time-varying and static distributions, show the potential of the proposed method for unsupervised learning.     

ALADIN: A scanner generator for incremental programming environments

A large number of scanner generators have been developed. Since they are restricted to the longest-match rule, they are unsuitable for an incremental environment. We present the ALADIN system, which is able to deliver more than a single token if required. Thus, an ambiguity may be passed to the calling instance. Beyond this ‘incremental feature’, ALADIN is a well-structured and easy-to-understand language. In contrast to existing systems, the desired behaviour of the generated scanners is completely specified explicitly. Thus, the specifications are more abstract than in other systems. A prototype implementation has shown that ALADIN-generated scanners have about the same performance as those generated by Lex.     

A robust analysis of adversarial attacks on federated learning environments

Federated Learning is a growing branch of Artificial Intelligence with the wide usage of mobile computing and IoT technologies. Since this technology uses distributed computing paradigm to do the learning part, most of the participating components are mobile devices and come outside the range of protection offered by a centralized system. As a result, several security issues such as data leakage, communication issues, poisoning, system manipulation via the backdoor, and so on arise with the usage of such a methodology. These sorts of attacks are categorized into various categories concerning their modus operandi. In this study, we review such attacks, namely poisoning attacks, inferencing attacks, their types, and working in a Federated Learning environment in detail. This study will give a precise idea of security issues faced in Federated Machine Learning and possible solutions.     

Spontaneous facial expression recognition: A robust metric learning approach

Spontaneous facial expression recognition is significantly more challenging than recognizing posed ones. We focus on two issues that are still under-addressed in this area. First, due to the inherent subtlety, the geometric and appearance features of spontaneous expressions tend to overlap with each other, making it hard for classifiers to find effective separation boundaries. Second, the training set usually contains dubious class labels which can hurt the recognition performance if no countermeasure is taken. In this paper, we propose a spontaneous expression recognition method based on robust metric learning with the aim of alleviating these two problems. In particular, to increase the discrimination of different facial expressions, we learn a new metric space in which spatially close data points have a higher probability of being in the same class. In addition, instead of using the noisy labels directly for metric learning, we define sensitivity and specificity to characterize the annotation reliability of each annotator. Then the distance metric and annotators' reliability is jointly estimated by maximizing the likelihood of the observed class labels. With the introduction of latent variables representing the true class labels, the distance metric and annotators' reliability can be iteratively solved under the Expectation Maximization framework. Comparative experiments show that our method achieves better recognition accuracy on spontaneous expression recognition, and the learned metric can be reliably transferred to recognize posed expressions.