In deep learning, understanding the decision-making processes of complex models is essential for advancing interpretability and trust in artificial intelligence systems. We introduce Causal Relational Attribution Graph (C-RAG), designed to deliver comprehensive, multi-perspective explanations of convolutional neural networks (CNNs) via a graph representation. C-RAG integrates gradient-based local attribution with global feature importance by constructing a graph-based representation that captures hierarchical feature inter-dependencies. In this framework, feature clusters are represented as graph nodes, and their interactions are quantified through combined localized and global attribution metrics, ensuring interpretable insights into model behavior. We evaluate C-RAG across diverse benchmark datasets (ImageNet, CIFAR-10, MNIST) and CNN architectures (ResNet18, VGG19, DenseNet201, LeNet), demonstrating significant advancements over state-of-the-art explainability methods in faithfulness, robustness, and computational efficiency. The proposed approach facilitates accurate spatial feature localization, robust dependency mapping, and efficient explanation generation, making it a valuable tool for critical applications such as medical imaging and autonomous systems. We provide a novel graph-based explainability framework, which bridges the gap between local and global interpretability, C-RAG addresses key limitations in existing methods, establishing a robust foundation for explainable AI in computer vision.

Abstract Deep Neural Networks (DNNs) have achieved state-of-the-art performance across various domains, yet their widespread adoption remains constrained by substantial computational and memory demands. While model pruning has emerged as a compelling strategy to address these challenges, existing methods often suffer from critical shortcomings: (1) non-selective neuron pruning which overlooks neuron activation dynamics, leading to the removal of critical neurons; (2) an inability to account for task-specific neuron importance, which causes accuracy degradation; and (3) failure to mitigate redundancy in neuron activations, resulting in suboptimal compression. In this work, we propose a novel Neuron Efficiency Metric (NEM), which integrates three key components—Neuron Activation Rate (NAR), Class-specific Activation Strength (CAS), and Neuron Overlap Index (NOI)—to address these limitations and guide a more effective pruning process. By iteratively evaluating the relevance of each neuron along these axes, NEM ensures selective and structured pruning that minimizes the retention of redundant or irrelevant neurons while preserving task-critical activations. The proposed method is tested on modern architectures using benchmark datasets such as MNIST and CIFAR-10, demonstrating a significant reduction in computational complexity while maintaining or even improving model performance. The results reveal that NEM achieves a higher degree of compression with minimal accuracy loss compared to conventional techniques.

Accurate position estimation is crucial for legged robots, the inertial odometry based on inertial measurement units (IMUs) is low-cost and easy to deploy, and inertial simultaneous localization and mapping (SLAM) methods can maintain stable performance on different robot platforms and degraded environments. In this paper, a single IMU is mounted on the robot foot to construct the inertial odometry without relying on kinematic modeling. The stance phase and optimal zero-velocity point (OZP) are detected through IMU outputs, and the once zero-velocity update is performed at each OZP to correct the drift of inertial odometry. Furthermore, the enhanced occupancy grid inertial SLAM framework is introduced. In the grid map update section, the motion control vectors derived from the inertial odometry are used with the Bresenham algorithm to calculate accessibility probabilities of grid cells, which addresses detail loss from grid map discretization and the imbalance in the accessibility probability resulting from repeated erroneous map updates. Experimental tests on the Unitree B1 robot demonstrate that our system provides further improvement in position estimation.

Abstract Convolutional neural networks have become prominent machine learning models, particularly in the realm of computer vision, due to their ability to predict and extract robust features from raw image data. CNNs, similar to other neural network models, undergo training via backpropagation, an iterative technique. However, the backpropagation algorithm has notable challenges, including slow convergence, susceptibility to local minima, and hypersensitivity to learning rates. These challenges not only impact the model’s accuracy but also make the training process computationally intensive. To address these limitations, We introduce a novel approach that trains the CNN classifier using a non-iterative learning method. The proposed approach involves automatic extraction of pertinent features from the raw-data, followed by the application of Gram–Schmidt process to decompose the feature matrix and determine classifier’s weights. The proposed method has shown enhanced predictive accuracy over state-of-the-art models when evaluated on two benchmark datasets, MNIST and CIFAR-10. The extensive experimentation using most cited pre-trained experiments validate the effectiveness of our proposed method.

Traffic light recognition is crucial for autonomous driving. While significant progress has been made in favorable conditions, recognition performance in night-time scenes remains a challenge. One straightforward approach is to apply enhancement methods that improve degraded images prior to object detection. However, since most enhancement methods are tailored for human perception; they may not consistently improve recognition accuracy for machine learning techniques. To address this, we propose an enhancement-guided framework for night-time traffic light recognition, called EG-TLR. EG-TLR consists of a residual denoising module (RDM) and a mixed attention traffic light detection module (MATLDM). The RDM reduces noise in degraded night-time images while preserving essential traffic light features by extracting sparsity information and performing context aggregation. The MATLDM improves feature extraction and recognition performance in complex night-time scenes by incorporating a shadow detection layer (SDL) and a mixed attention module (MAM). Moreover, to address the lack of a dedicated night-time traffic light dataset, we construct the Night-TL dataset utilizing publicly available images. Extensive experiments on Night-TL and LISA datasets demonstrate that EG-TLR achieves an AP50 of 79.83% and an AP50:95 of 36.62%, with an inference speed of 5.9 ms and 16.9 GFLOPs, outperforming other state-of-the-art methods. Furthermore, ablation studies and visualization results validate the effectiveness of our proposed method. The Night-TL dataset can be downloaded from: https://github.com/feiqinaqian/Night-TL-dataset.

Deep-learning models are popular machine learning models that have gained their popularity in various fields of computer vision, natural language processing etc, due to their excellent predictive accuracy and ability to automatically extract good features from raw input data. Though these models have these significant advantages, most deep learning models are notoriously black-box models. The features learned by these models are not explainable. The need to explain these predictions is important in many high stakes fields. In this work, we propose a method that uses convolutional layers like in Convolutional Neural Networks (CNNs) that automatically learns features from raw data. In our proposed method we extract feature representation that is unique to each target class in the given data. Visualizations of this unique representation explain the feature learned by the model for a specific target class. To classify an input data we use two approaches. The first approach uses the similarity scores to match the features learned for a particular instance with the extracted target class-specific features to make prediction. The second approach uses an intrinsically explainable model like logistic regression to make predictions. Thus our proposed method in addition to having good predictive accuracy, identifies features that are class-specific. The proposed method achieved an accuracy of 99.31 % on MNIST data set, 90.45% on CIFAR-10 data set and 96.45% accuracy on Assira data set.

This paper addresses the blind recovery of the parity check matrix of an (n, k) linear block code over noisy channels by proposing a fast recovery scheme consisting of 3 parts. Firstly, this scheme performs initial error position detection among the received codewords and selects the desirable codewords. Then, this scheme conducts Gaussian elimination (GE) on a k-by-k full-rank matrix and uses a threshold and the reliability associated to verify the recovered dual words, aiming to improve the reliability of recovery. Finally, it performs decoding on the received codewords with partially recovered dual words. These three parts can be combined into different schemes for different noise level scenarios. The GEV that combines Gaussian elimination and verification has a significantly lower recovery failure probability and a much lower computational complexity than an existing Canteaut-Chabaud-based algorithm, which relies on GE on n-by-n full-rank matrices. The decoding-aided recovery (DAR) and error-detection-&-codeword-selection-&-decoding-aided recovery (EDCSDAR) schemes can improve the code recovery performance over GEV for high noise level scenarios, and their computational complexities remain much lower than the Canteaut-Chabaud-based algorithm.

This work investigated the use of an ensemble deep random vector functional link (edRVFL) network for electroencephalogram (EEG)-based driver fatigue recognition. Against the low feature learning capability of the edRVFL network from raw EEG signals, two strategies were exploited in this work. Specifically, the first one was to exploit the advantages of the feature extractor module in CNNs, i.e., use CNN features as the input of the edRVFL network. The second one was to improve the feature learning capability of the edRVFL network. An enhanced edRFVL network named FGloWD-edRVFL was proposed, in which four enhancements were implemented, including random forest-based Feature selection, Global output layer, Weighting and entropy-based Dynamic ensemble. The proposed FGloWD-edRVFL network was evaluated on the challenging cross-subject driver fatigue recognition tasks. The results indicated that the proposed model could boost the recognition performance, significantly outperforming all strong baselines. The step-wise analysis further demonstrated the effectiveness of the proposed enhancements in the edRVFL network.

Randomized neural networks (RNNs) have shown outstanding performance in many different fields. The superiority of having fewer training parameters and closed-form solutions makes them popular in small datasets analysis. However, automatically decoding raw electroencephalogram (EEG) data using RNNs is still challenging in EEG-based passive brain–computer interface (pBCI) classification tasks. Models with the high-dimension input of EEG may suffer from overfitting and the intrinsic characteristics of non-stationary, high-level noises and subject variability could limit the generation of distinctive features in the hidden layers. To address these problems in EEG-based pBCI tasks, this work proposes a spectral-ensemble deep random vector functional link (SedRVFL) network that focuses on feature learning in the frequency domain. Specifically, an unsupervised feature-refining (FR) block is proposed to improve the low feature learning capability in RNNs. Moreover, a dynamic direct link (DDL) is performed to further complement the frequency information. The proposed model has been evaluated on a self-collected dataset as well as a public driving dataset. The cross-subject classification results obtained demonstrated its effectiveness. This work offers a new solution for EEG decoding, i.e., using optimized RNNs for decoding complex raw EEG data and boosting the classification performance of EEG-based pBCI tasks.

The emergence of the cloud platform with substantial resources to offer on-demand instigated the researchers to migrate the scientific workflows to the cloud environment. The scheduling of workflows with diverse QoS parameters is not a trivial task, but an NP-Complete problem. Several heuristics for QoS constrained workflows have been investigated. However, most of them focus only on time and cost and do not guarantee high resource utilization. The scheduling of the workflow tasks over the minimum cloud resources under the defined time limit is a grave concern. In this article, an algorithm named MFGA (Modified Fuzzy Adaptive Genetic Algorithm) has been formulated to minimize the makespan and improve resource utilization under both deadline and budget constraints. A fuzzy logic controller has also been devised to control the crossover and mutation rates that prevent MFGA from getting stuck in a local optimum. MFGA has a novel crossover technique that adds the fittest solutions in the population. Additionally, a new mutation technique has also been introduced, which minimizes the makespan and increases the reusability of the resources. The simulation experiments with the real workflows show that the proposed MFGA outperforms other state-of-the-art algorithms.

Since situation awareness (SA) plays an important role in many fields, the measure of SA is one of the most concerning problems. Using physiological signals to evaluate SA is becoming a popular research topic because of their advantages of non-intrusiveness and objectivity. However, previous studies mainly exploited the use of single physiological signals such as electroencephalogram (EEG) or eye tracking. The multi-modal SA recognition is still a research gap. Therefore, this work conducts a pilot study to investigate SA recognition by using two modalities: EEG and eye tracking data. Specifically, an optimized Stroop test that is more compatible with the definition of SA was used to induce different states of SA and collect physiological data. Furthermore, a random vector functional link-based stacking (RVFL-S) model was proposed to perform the multi-modal SA recognition. Experiment results showed that using the combination of EEG and eye tracking data can boost the performance of SA recognition. Moreover, the proposed RVFL-S model can effectively integrate the classification information from two modalities. It showed better performance than baseline methods, achieving 77.62% leave-one-subject-out (LOSO) average accuracy. This was around 5% improvement compared with the baseline classification models with input of only one modality. This pilot study demonstrated that the use of multi-modality is a potential strategy for SA recognition.

While deep learning technologies are now capable of generating realistic images confusing humans, the research efforts are turning to the synthesis of images for more concrete and application-specific purposes. Facial image generation based on vocal characteristics from speech is one of such important yet challenging tasks. It is the key enabler to influential use cases of image generation, especially for business in public security and entertainment. Existing solutions to the problem of speech2face renders limited image quality and fails to preserve facial similarity due to the lack of quality dataset for training and appropriate integration of vocal features. In this paper, we investigate these key technical challenges and propose Speech Fusion to Face, or SF2F in short, attempting to address the issue of facial image quality and the poor connection between vocal feature domain and modern image generation models. By adopting new strategies on data model and training, we demonstrate dramatic performance boost over state-of-the-art solution, by doubling the recall of individual identity, and lifting the quality score from 15 to 19 based on the mutual information score with VGGFace classifier.

Deep learning for electroencephalogram-based classification is confronted with data scarcity, due to the time-consuming and expensive data collection procedure. Data augmentation has been shown as an effective way to improve data efficiency. In addition, contrastive learning has recently been shown to hold great promise in learning effective representations without human supervision, which has the potential to improve the electroencephalogram-based recognition performance with limited labeled data. However, heavy data augmentation is a key ingredient of contrastive learning. In view of the limited number of sample-based data augmentation in electroencephalogram processing, three methods, performance-measure-based time warp, frequency noise addition and frequency masking, are proposed based on the characteristics of electroencephalogram signal. These methods are parameter learning free, easy to implement, and can be applied to individual samples. In the experiment, the proposed data augmentation methods are evaluated on three electroencephalogram-based classification tasks, including situation awareness recognition, motor imagery classification and brain-computer interface steady-state visually evoked potentials speller system. Results demonstrated that the convolutional models trained with the proposed data augmentation methods yielded significantly improved performance over baselines. In overall, this work provides more potential methods to cope with the problem of limited data and boost the classification performance in electroencephalogram processing.

Acoustic resolution photoacoustic micros- copy (AR-PAM) can achieve deeper imaging depth in biological tissue, with the sacrifice of imaging resolution compared with optical resolution photoacoustic microscopy (OR-PAM). Here we aim to enhance the AR-PAM image quality towards OR-PAM image, which specifically includes the enhancement of imaging resolution, restoration of micro-vasculatures, and reduction of artifacts. To address this issue, a network (MultiResU-Net) is first trained as generative model with simulated AR-OR image pairs, which are synthesized with physical transducer model. Moderate enhancement results can already be obtained when applying this model to in vivo AR imaging data. Nevertheless, the perceptual quality is unsatisfactory due to domain shift. Further, domain transfer learning technique under generative adversarial network (GAN) framework is proposed to drive the enhanced image's manifold towards that of real OR image. In this way, perceptually convincing AR to OR enhancement result is obtained, which can also be supported by quantitative analysis. Peak Signal to Noise Ratio (PSNR) and Structural Similarity Index (SSIM) values are significantly increased from 14.74 dB to 19.01 dB and from 0.1974 to 0.2937, respectively, validating the improvement of reconstruction correctness and overall perceptual quality. The proposed algorithm has also been validated across different imaging depths with experiments conducted in both shallow and deep tissue. The above AR to OR domain transfer learning with GAN (AODTL-GAN) framework has enabled the enhancement target with limited amount of matched in vivo AR-OR imaging data.

For EEG-based drowsiness recognition, it is desirable to use subject-independent recognition since conducting calibration on each subject is time-consuming. In this paper, we propose a novel Convolutional Neural Network (CNN)-Long Short-Term Memory (LSTM) model for subject-independent drowsiness recognition from single-channel EEG signals. Different from existing deep learning models that are mostly treated as black-box classifiers, the proposed model can “explain” its decisions for each input sample by revealing which parts of the sample contain important features identified by the model for classification. This is achieved by a visualization technique by taking advantage of the hidden states output by the LSTM layer. Results show that the model achieves an average accuracy of 72.97% on 11 subjects for leave-one-out subject-independent drowsiness recognition on a public dataset, which is higher than the conventional baseline methods of 55.42%-69.27%, and state-of-the-art deep learning methods. Visualization results show that the model has discovered meaningful patterns of EEG signals related to different mental states across different subjects.

Extracting intra- and inter-subject parameters from Electroencephalogram (EEG) representing different Situation Awareness (SA) status is a critical challenge for objective SA recognition. Most of the existing work focuses on the subject-dependent classification that applies power spectrum density (PSD) features. In this paper, we propose a novel spectral-spatial (S-S) model for cross-subject fatigue-related SA recognition. The S-S model not only considers the biological topology across different brain regions to capture both local and global relations among different EEG channels, but also extracts spectral features for each EEG channel. Specifically, we firstly model the topological structure of EEG channels via an adjacency matrix which is built based on the Euclidean distance between EEG channels. Then, the graph convolution operation is employed to perform the neighbourhood aggregation for extracting spatial features. We test our model on a public dataset collected during driver’s task performance. The subject-independent performance of the model is explored. Results demonstrate (1) the superior performance of our model compared with the state-of-the-art models on SA recognition from EEG signals. Specifically, our S-S model achieves 70.6% accuracy which is higher than traditional machine learning methods by 2.7%-6.8% and deep learning methods by 10.3%-11.6%; (2) EEG signal at the occipital region can better reflect the change of SA.

Affective brain-computer interface (aBCI) introduces personal affective factors into human-computer interactions, which could potentially enrich the user's experience during the interaction with a computer. However, affective neural patterns are volatile even within the same subject. To maintain satisfactory emotion recognition accuracy, the state-of-the-art aBCIs mainly tailor the classifier to the subject-of-interest and require frequent re-calibrations for the classifier. In this paper, we demonstrate that the recognition accuracy of aBCIs deteriorates when re-calibration is ruled out during the long-term usage for the same subject. Then, we propose a stable feature selection method to choose the most stable affective features, for mitigating the accuracy deterioration to a lesser extent and maximizing the aBCI performance in the long run. We validate our method on a dataset comprising six subjects' EEG data collected during two sessions per day for each subject for eight consecutive days.

For practical applications, it is desirable for a trained classification system to be independent of task and/or subject. In this study, we show one-way transfer between two independent EEG workload datasets: from a large multitasking dataset with 48 subjects to a second Stroop test dataset with 18 subjects. This was achieved with a classification system trained using sparse encoded representations of the decomposed wavelets in the alpha, beta and theta power bands, which learnt a feature representation that outperformed benchmark power spectral density features by 3.5%. We also explore the possibility of enhancing performance with the utilization of domain adaptation techniques using transfer component analysis (TCA), obtaining 30.0% classification accuracy for a 4-class cross dataset problem.

Building behavior profiles of Android applications (apps) with holistic, rich and multi-view information (e.g., incorporating several semantic views of an app such as API sequences, system calls, etc.) would help catering downstream analytics tasks such as app categorization, recommendation and malware analysis significantly better. Towards this goal, we design a semisupervised Representation Learning (RL) framework named apk2vec to automatically generate a compact representation (aka profile/embedding) for a given app. More specifically, apk2vec has the three following unique characteristics which make it an excellent choice for large-scale app profiling: (1) it encompasses information from multiple semantic views such as API sequences, permissions, etc., (2) being a semi-supervised embedding technique, it can make use of labels associated with apps (e.g., malware family or app category labels) to build high quality app profiles, and (3) it combines RL and feature hashing which allows it to efficiently build profiles of apps that stream over time (i.e., online learning). The resulting semi-supervised multi-view hash embeddings of apps could then be used for a wide variety of downstream tasks such as the ones mentioned above. Our extensive evaluations with more than 42,000 apps demonstrate that apk2vec's app profiles could significantly outperform state-of-the-art techniques in four app analytics tasks namely, malware detection, familial clustering, app clone detection and app recommendation.

Summary The increasingly prevalent use of mobile devices has raised the popularity of mobile applications. Therefore, automated testing of mobile applications has become an extremely important task. However, it is still a challenge to automatically generate tests with high coverage for mobile applications due to their specific nontrivial structure and the highly interactive nature of graphical user interfaces (GUIs). In this paper, we propose a novel automated GUI testing technique for mobile applications, namely, Mobolic. In this approach, tests with high coverage are automatically generated and executed by combining the online testing technique and customated input generation. Employing the online testing technique, Mobolic systematically explores the app GUI without falling in a loop. It generates relevant events “on the fly” that are followed by an immediate execution. In addition, involving the customated input generation, Mobolic automatically generates relevant user inputs such as user‐predefined, concrete, or random ones. We implemented Mobolic and evaluated its performance on 10 real‐world open‐source Android applications. Our experimental results show the effectiveness and efficiency of Mobolic in terms of achieved code coverage and overall exercising time.

An insider threat scenario refers to the outcome of a set of malicious activities caused by intentional or unintentional misuse of the organization's systems, networks, data, and resources. Prevention of insider threat is difficult, since trusted partners of the organization are involved in it, who have authorized access to these confidential/sensitive resources. The state-of-the-art research on insider threat detection mostly focuses on developing unsupervised behavioral anomaly detection techniques with the objective of finding out anomalousness or abnormal changes in user behavior over time. However, an anomalous activity is not necessarily malicious that can lead to an insider threat scenario. As an improvement to the existing approaches, we propose a technique for insider threat detection from time-series classification of user activities. Initially, a set of single-day features is computed from the user activity logs. A time-series feature vector is next constructed from the statistics of each single-day feature over a period of time. The label of each time-series feature vector (whether malicious or nonmalicious) is extracted from the ground truth. To classify the imbalanced ground-truth insider threat data consisting of only a small number of malicious instances, we employ a cost-sensitive data adjustment technique that undersamples the nonmalicious class instances randomly. As a classifier, we employ a two-layered deep autoencoder neural network and compare its performance with other popularly used classifiers: random forest and multilayer perceptron. Encouraging results are obtained by evaluating our approach using the CMU Insider Threat Data, which is the only publicly available insider threat data set consisting of about 14-GB web-browsing logs, along with logon, device connection, file transfer, and e-mail log files. We observe that both deep autoencoder and random forest classifiers classify the dataadjusted time-series feature set with high precision, recall, and f-score. Although multilayer perceptron has a high recall, it suffers from a lower precision and f-score compared to the other two classifiers.

Spectral band power features are one of the most widely used features in the studies of electroencephalogram (EEG)-based emotion recognition. The power spectral density of EEG signals is partitioned into different bands such as delta, theta, alpha and beta band etc. Though based on neuroscientific findings, the partition of frequency bands is somewhat on an ad-hoc basis, and the definition of frequency ranges of the bands of interest can vary between studies. On the other hand, it is also arguable that one definition of power bands could perform equally well on all subjects. In this paper, we propose to use autoencoder to automatically learn from each subject the salient frequency components from power spectral density estimated as periodogram by Fast Fourier Transform (FFT). We propose a network architecture especially for EEG feature extraction, one that adopts hidden unit clustering with added pooling neuron per cluster. The classification accuracy with features extracted by our proposed method is benchmarked against that with standard power features. Experimental results show that our proposed feature extraction method achieves accuracy ranging from 44% to 59% for three-emotion classification. We also see a 4-20% accuracy improvement over standard band power features.

In 2006 Zhang and Zhou proposed a multilabel classification model based on the MLP network, which was subsequently improved by Grodzicki et al. This paper further improves both these approaches by introducing a scaling parameter responsible for maintaining a balance between the impacts of particular components of the MLP's error function in the training process. The newly-proposed parameter is autonomously fine-tuned by the system in the nested cross validation process. The proposed approach is tested on a set of well-established benchmarks and demonstrates its superiority over the baseline methods for 16 different error measures used in the experiments. Furthermore, the method proves competitive to 12 other state-of-the-art machine learning approaches which are used for further comparisons. In the combined score composed of ranking positions for all benchmarks and all error functions, the proposed neural network system gains the leading position among all tested methods.

This special issue of Concurrency and Computation: Practice & Experience (CCPE) aims to publish some excellent works presented in the 11th International Conference on Natural Computation (ICNC'15) and the 12th International Conference on Fuzzy Systems and Knowledge Discovery (FSKD'15), jointly held during August 15-17, 2015, at Zhangjiajie, China. ICNC-FSKD is a premier international forum for scientists and researchers to present the state-of-the-art of data mining and intelligent methods inspired from nature, particularly biological, linguistic, and physical systems, with applications to computers, circuits, systems, control, communications, and more. ICNC-FSKD 2015 totally received 738 manuscripts from over 40 countries. More than 500 people from universities, institutes, and companies attended this conference. Through two round peer to peer reviews, 9 papers are selected after patiently extending and revising manuscripts to meet the comments/suggestions of the referees and responding to meet all the requirements of the journal. The topics covered in this special issue include Cloud Computing 1, 2, Parallel and Distributed Processing 3-5, Scheduling of Resources 6, 7, and other applications in High-Performance Computing for Natural Computation and Knowledge Discovery 8, 9, It is hoped that this CCPE issue will make a good reference material and be of great use for readers in Computer Science, System Engineering, Mathematics, and Artificial Intelligence, etc.

Android applications typically contain multiple third-party libraries and recent studies have shown that the presence of third-party libraries may introduce privacy risks and security threats. Furthermore, researchers have reported the importance of considering the third-party libraries for their program analysis tasks. A reason being that the presence of third-party libraries may dilute the features and affect the accuracy of their results. Existing literature typically employs a whitelist to exclude the third-party libraries from their analysis in order to achieve accurate results. However, these whitelists are generally incomplete and weak against the renaming obfuscation technique that is commonly employed in Android applications. In this paper, we propose LibSift, a tool to automatically detect third-party libraries in Android applications. LibSift detects third-party libraries based on package dependencies that are resilient to most common obfuscations. The evaluation results not only indicate that LibSift can detect third-party libraries accurately and effectively, but also show that LibSift can detect even the less popular libraries that are not detected by two of the state-of-the-art approaches.

Neuroscience-based or neuroscience-informed design is a new application area of Brain-Computer Interaction (BCI). It takes its roots in study of human well-being in architecture, human factors study in engineering and manufacturing including neuroergonomics. In traditional human factors studies and/or well-being study, mental workload, stress, and emotion are obtained through questionnaires that are administered upon completion of some task and/or the whole experiment. Recent advances in BCI research allow for using Electroencephalogram (EEG) based brain state recognition algorithms to assess the interaction between brain and human performance. We propose and develop an EEG-based system CogniMeter to monitor and analyze human factors measurements of newly designed software/hardware systems and/or working places. Machine learning techniques are applied to the EEG data to recognize levels of mental workload, stress and emotions during each task. The EEG is used as a tool to monitor and record the brain states of subjects during human factors study experiments. We describe two applications of CogniMeter system: human performance assessment in maritime simulator and EEG-based human factors evaluation in Air Traffic Control (ATC) workplace. By utilizing the proposed EEG-based system, true understanding of subjects working patterns can be obtained. Based on the analyses of the objective real time EEG-based data together with the subjective feedback from the subjects, we are able to reliably evaluate current systems/hardware and/or working place design and refine new concepts and design of future systems.

Emotion recognition is an integral part of affective computing. An affective brain-computer-interface (BCI) can benefit the user in a number of applications. In most existing studies, EEG (electroencephalograph)-based emotion recognition is explored in a classificatory manner. In this manner, human emotions are discretized by a set of emotion labels. However, human emotions are more of a continuous phenomenon than discrete. A regressive approach is more suited for continuous emotion recognition. Few studies have looked into a regressive approach. In this study, we investigate a portfolio of EEG features including fractal dimension, statistics and band power. Support vector regression (SVR) is employed in this study to estimate subject's valence level by means of different features under two evaluation schemes. In the first scheme, a SVR is constructed with full training resources, whereas in the second scheme, a SVR only receives minimal training resources. MAE (mean absolute error) averages of 0.74 and 1.45 can be achieved under the first and the second scheme, respectively, by fractal feature. The advantages of a regressive approach over classificatory approach lie in continuous emotion recognition and the possibility to reduce training resources to minimal level.

The individual alpha peak frequency (IAPF) is an important biological indicator in Electroencephalogram (EEG) studies, with many research publications linking it to various cognitive functions. In this paper, we propose novel Power Spectral Density (PSD) alpha features based on IAPF to classify 2 and 4 levels of EEG multitasking workload data. When optimized IAPF was considered, a 1.55% and 1.56% increase in average accuracy for 48 subjects' data, with 35 and 33 subjects showing improvement was observed for 2 and 4 class cases respectively. This trend suggests that individual specific features are able to improve classification performance compared to generalized features for subject dependent cases. The proposed features, which incorporates the biological meaning of the IAPF and provides subject specific information, can be considered as a viable alternative to the general alpha power feature when designing novel subject dependent feature sets for BCI workload recognition applications.

Multitasking is a prevalent phenomenon in our daily lives. Certain occupations, especially in the aviation industry, consider proficient multitasking as a key skill set in their hiring process for pilot or air traffic controller candidates. There is a growing interest in the testing and training of the multitasking ability, with in house software or commercial psychological products, usually implemented in a static task battery format. In this paper, we propose a 3D game, Multitask In Neurofeedback Driving (MIND) for training and testing of the multitasking ability. The game is developed using the Unreal 3 game engine and incorporates neurofeedback, a technique used in the training of human cognitive abilities, to further enhance the potential benefits of the training procedure. The tasks used in the multitasking condition are inspired by various psychological tests and implemented in a manner that attempts to simulate the general cognitive processes required for multitasking while driving a vehicle or piloting an aircraft. The game comes in three variants, single task condition, multitasking condition and multitasking with neurofeedback condition, for the purpose of validating the training outcomes in future studies.

The difference in cognitive abilities of humans could be assessed by indexes extracted from EEG. In this paper, we propose and implement an experiment with 60 subjects to study how cognitive abilities can be identified through EEG. We analyzed parameters of the individual frequency band that can be used for prediction of cognitive abilities of subjects. In the experiment, the subjects performed cognitive tests with EEG recording done prior to the tests. Different patterns of alpha band activity are proven to be indicators of cognitive abilities and performances. Our hypothesis is that cognitive abilities can be predicted based on EEG measurements. The results of analysis of the experiment show significant correlation between subjects' cognitive abilities assessed by the tests and the EEG measurements.

The blooming mobile smart phone device industry has attracted a large number of application developers. However, due to the availability of reverse engineering tools for Android applications, it also caught the attention of plagiarists and malware writers. In recent years, application cloning has become a serious threat to the Android market. In previous work, mobile application clone detection mainly focuses on code-based analysis. Such an approach lacks resilient to advanced obfuscation techniques. Their efficiency is also questionable, as billions of opcodes need to be processed for cross-market clone detection. In this paper, we propose a novel technique of detecting Android application clones based on the analysis of user interface (UI) information collected at runtime. By leveraging on the multiple entry points feature of Android applications, the UI information can be collected easily without the need to generate relevant inputs and execute the entire application. Another advantage of our technique is obfuscation resilient since semantics preserving obfuscation technique do not affect runtime behaviors. We evaluated our approach on a set of real-world dataset and it has a low false positive rate and false negative rate. Furthermore, the results also show that our approach is effective in detecting different types of repackaging attacks.

Everyone experiences stress in life. Moderate stress can be beneficial to human, however, excessive stress is harmful to the health. To monitor stress, different methods can be used. In this work, an algorithm for stress level recognition from Electroencephalogram (EEG) is proposed. To validate the algorithm, an experiment is designed and carried out with 9 subjects. A Stroop colour-word test is used as a stressor to induce 4 levels of stress, and the EEG data are recorded during the experiment. Different feature combinations and classifiers are proposed and analyzed. By combining fractal dimension and statistical features and using Support Vector Machine (SVM) as the classifier, four levels of stress can be recognized with an average accuracy of 67.06%, three levels of stress can be recognized with an accuracy of 75.22%, and two levels of stress can be recognized with an accuracy of 85.71%. The algorithm is integrated into the system CogniMeter for stress state monitoring. Stress level of the user is visualized on the meter in real time. The system can be applied for stress monitoring of air traffic controllers, operators, etc.

Stability of algorithms is very important for electroencephalogram (EEG) based applications. Stable features should exhibit consistency among repeated measurements of the same subject. Previously, power features were reported to be one of the most stable EEG features in medical application. In this paper, stability of features in emotion recognition algorithms is studied. Our hypothesis is that the most stable features give the best intra-subject accuracy across different days in real-time emotion recognition algorithm. An experiment to induce 4 emotions such as pleasant, happy, frightened, and angry is designed and carried out in 8 consecutive days (two sessions per day) for 4 subjects to record EEG data. A novel real-time subject dependent algorithm with the most stable features is proposed and implemented. The algorithm needs just one training for each subject. The training results can be used in real-time emotion recognition applications without re-training with the adequate accuracy. The proposed algorithm is integrated with a real-time application "Emotional Avatar".

This paper improves stock market prediction based on genetic algorithms (GA) and wavelet neural networks (WNN) and reports significantly better accuracies compared to existing approaches to stock market prediction, including the hierarchical GA (HGA) WNN. Specifically, we added information such as trading volume as inputs and we used the Morlet wavelet function instead of Morlet-Gaussian wavelet function in our prediction model. We also employed a smaller number of hidden nodes in WNN compared to other research work. The prediction system is tested using Shenzhen Composite Index data.

In this paper, an intelligent model for analyzing and predicting cloud computing resource utilization is proposed to enhance on-demand services in cloud computing systems. The model is with the capability to discover active users and mine the system storage utilization patterns. This model is also with learning capabilities to adapt the dynamics in the cloud computing platform by capturing changing patterns of system storage utilization, and it employs data mining means for computing the practical model to be used for prediction and providing inputs for intelligent management in the on-demand cloud computing system. We have evaluated the proposed analysis and prediction model in a cloud computing platform. High prediction accuracies of 95% and 86% have been achieved in 1-day ahead and 7-day ahead system utilization prediction, respectively.

Relevance feedback (RF) schemes have been widely designed to improve the performance of content-based image retrieval. Despite the success, it is not appropriate to require the user to label a large number of samples in RF. Collaborative image retrieval (CIR) aims to reduce the labeling efforts of the user by resorting to the auxiliary information. Support vector machine (SVM) active learning can select ambiguous samples as the most informative ones for the user to label with the help of the optimal hyperplane of SVM, and thus alleviate the labeling efforts of conventional RF. However, the optimal hyperplane of SVM is usually unstable and inaccurate with small-sized training data, and this is always the case in image retrieval since the user would not like to label a large number of feedback samples and cannot label each sample accurately all the time. In this paper, we propose a novel active learning method, i.e., geometric optimum experimental design (GOED), to select multiple representative samples in the database as the most informative ones for the user to label. Especially, GOED can alleviate the small-sized training data problem by leveraging the geometric structure of unlabeled samples in the reproducing kernel Hilbert space and thus further enhance the performance of image retrieval. Different from the conventional manifold regularization framework, the new method can effectively select the most informative samples for the user to label in image retrieval. By minimizing the expected average prediction variance on the test data, GOED has a clear geometric interpretation to select a set of the most representative samples in the database iteratively with the global optimum. Compared with the popular SVM active learning, our method is label-independent and can effectively avoid various potential problems caused by insufficient and inexactly labeled samples in RF, and is more appropriate and useful for image retrieval. Extensive experiments on both synthetic datasets and a real-world image database have been conducted to show the advantages of the proposed GOED for CIR.

For computer vision-based inspection of electronic chips or dies in semiconductor production lines, we propose a new method to effectively and efficiently detect defects in images. Different from the traditional methods that compare the image of each test chip or die with the template image one by one, which are sensitive to misalignment between the test and template images, a collection of multiple test images are used as the input image for processing simultaneously in our method with two steps. The first step is to obtain salient regions of the whole collection of test images, and the second step is to evaluate local discrepancy between salient regions in test images and the corresponding regions in the defect-free template image. To be more specific, in the first step of our method, phase-only Fourier transform (POFT), which is computationally efficient for online applications in industry, is used for saliency detection. We provide the theoretical justification for POFT to be effective to attenuate the normal regions and amplify the defects in multiple test images, which are usually arranged in a matrix format in industrial practice. By comparing with four other popular methods, the proposed algorithm can efficiently accommodate small variations (inevitable in practice) in test chips or dies, such as the spatial misalignments and product variations. Experimental results on a large-scale database including 1073 images, 94 of which are defective, show that our method performs much better than the other methods in terms of precision, recall, and F-measure.

Conventional content-based image retrieval (CBIR) systems with the Euclidean distance metric in a high-dimensional visual feature space usually cannot achieve satisfactory performance due to the semantic gap. Relevance feedback (RF) has been introduced as a powerful tool to involve the user in the system to improve the performance of CBIR. Despite the success, an on-line learning task can be tedious and boring for the user. Various schemes have been proposed to exploit the RF log data to further enhance the performance of CBIR. In this paper, we propose a semantic subspace learning (SSL) method to exploit the RF log data with contextual information for an image retrieval task. Different from conventional subspace learning approaches, our method can directly learn a semantic concept subspace from the RF log data with contextual information without using any class label information. We show that the performance of the image retrieval task can be significantly improved in the low-dimensional semantic concept subspace. Extensive experiments on a real-world image database demonstrate the effectiveness of the proposed scheme in improving the performance of CBIR by exploiting the RF log data.

Content-based image retrieval (CBIR) has attracted substantial attention during the past few years for its potential applications. To bridge the gap between low level visual features and high level semantic concepts, various relevance feedback (RF) or interactive re-ranking (IR) schemes have been designed to improve the performance of a CBIR system. In this paper, we propose a novel subspace learning based IR scheme by using a graph embedding framework, termed Laplacian Regularized Subspace Learning (LRSL). The LRSL method can model both within-class compactness and between-class separation by specially designing an intrinsic graph and a penalty graph in the graph embedding framework, respectively. In addition, LRSL can share the popular assumption of the biased discriminant analysis (BDA) for IR but avoid the singular problem in BDA. Extensive experimental results have shown that the proposed LRSL method is effective for reducing the semantic gap and targeting the intentions of users for an image retrieval task.

With many potential practical applications, content-based image retrieval (CBIR) has attracted substantial attention during the past few years. A variety of relevance feedback (RF) schemes have been developed as a powerful tool to bridge the semantic gap between low-level visual features and high-level semantic concepts, and thus to improve the performance of CBIR systems. Among various RF approaches, support-vector-machine (SVM)-based RF is one of the most popular techniques in CBIR. Despite the success, directly using SVM as an RF scheme has two main drawbacks. First, it treats the positive and negative feedbacks equally, which is not appropriate since the two groups of training feedbacks have distinct properties. Second, most of the SVM-based RF techniques do not take into account the unlabeled samples, although they are very helpful in constructing a good classifier. To explore solutions to overcome these two drawbacks, in this paper, we propose a biased maximum margin analysis (BMMA) and a semisupervised BMMA (SemiBMMA) for integrating the distinct properties of feedbacks and utilizing the information of unlabeled samples for SVM-based RF schemes. The BMMA differentiates positive feedbacks from negative ones based on local analysis, whereas the SemiBMMA can effectively integrate information of unlabeled samples by introducing a Laplacian regularizer to the BMMA. We formally formulate this problem into a general subspace learning task and then propose an automatic approach of determining the dimensionality of the embedded subspace for RF. Extensive experiments on a large real-world image database demonstrate that the proposed scheme combined with the SVM RF can significantly improve the performance of CBIR systems.

Content-Based Image Retrieval (CBIR) has attracted substantial attention during the past few years for its potential practical applications to image management. A variety of Relevance Feedback (RF) schemes have been designed to bridge the semantic gap between the low-level visual features and the high-level semantic concepts for an image retrieval task. Various Collaborative Image Retrieval (CIR) schemes aim to utilize the user historical feedback log data with similar and dissimilar pairwise constraints to improve the performance of a CBIR system. However, existing subspace learning approaches with explicit label information cannot be applied for a CIR task, although the subspace learning techniques play a key role in various computer vision tasks, e.g., face recognition and image classification. In this paper, we propose a novel subspace learning framework, i.e., Conjunctive Patches Subspace Learning (CPSL) with side information, for learning an effective semantic subspace by exploiting the user historical feedback log data for a CIR task. The CPSL can effectively integrate the discriminative information of labeled log images, the geometrical information of labeled log images and the weakly similar information of unlabeled images together to learn a reliable subspace. We formally formulate this problem into a constrained optimization problem and then present a new subspace learning technique to exploit the user historical feedback log data. Extensive experiments on both synthetic data sets and a real-world image database demonstrate the effectiveness of the proposed scheme in improving the performance of a CBIR system by exploiting the user historical feedback log data.

Biased discriminant analysis (BDA) is one of the most promising relevance feedback (RF) approaches to deal with the feedback sample imbalance problem for content-based image retrieval (CBIR). However, the singular problem of the positive within-class scatter and the Gaussian distribution assumption for positive samples are two main obstacles impeding the performance of BDA RF for CBIR. To avoid both of these intrinsic problems in BDA, in this paper, we propose a novel algorithm called generalized BDA (GBDA) for CBIR. The GBDA algorithm avoids the singular problem by adopting the differential scatter discriminant criterion (DSDC) and handles the Gaussian distribution assumption by redesigning the between-class scatter with a nearest neighbor approach. To alleviate the overfitting problem, GBDA integrates the locality preserving principle; therefore, a smooth and locally consistent transform can also be learned. Extensive experiments show that GBDA can substantially outperform the original BDA, its variations, and related support-vector-machine-based RF algorithms.

This paper proposes an intelligent trading system using support vector regression optimized by genetic algorithms (SVR-GA) and multilayer perceptron optimized with GA (MLP-GA). Experimental results show that both approaches outperform conventional trading systems without prediction and a recent fuzzy trading system in terms of final equity and maximum drawdown for Hong Kong Hang Seng stock index.

In recent years, some emerging and reemerging infectious diseases have grown into global health threats due to high human mobility. It is important to have intervention plans for containing the spread of such infectious diseases. Among various intervention strategies, screening infected people is an efficient way for evaluating the infection scale and controlling the spread of infectious diseases. Considering the cost in manpower and limited screening machines available, we face to challenges for selecting the optimal nodes (sites) in order to obtain better screening and control effects. In this paper, particle swarm optimization technique is used to determine key nodes for controlling infectious disease spread, through evaluating the number of people captured at each key node. The research example is shown on evaluating the screening control over train stations in Singapore. The optimization algorithm and control concept can be easily extended to large-scale infectious disease control in other kinds of key nodes and in other geographical regions. The selection for optimal control set of the multi objective optimization problem is done using particle swarm optimization. Numerical simulation shows the effectiveness of the proposed algorithm.

We present a method to compute the delay constrained multicast routing tree by employing chaotic neural networks. Experimental result shows that the noisy chaotic neural network (NCNN) provides optimal solution more often compared to the transiently chaotic neural network (TCNN) and the Hopfield neural network (HNN). Furthermore, compared with the bounded shortest multicast algorithm (BSMA), the noisy chaotic neural network is able to find multicast trees with lower cost.

This comment letter points out that the essence of the "extreme learning machine (ELM)" recently appeared has been proposed earlier by Broomhead and Lowe and Pao , and discussed by other authors. Hence, it is not necessary to introduce a new name "ELM."

In this paper, we argue that for a C-class classification problem, C 2-class classifiers, each of which discriminating one class from the other classes and having a characteristic input feature subset, should in general outperform, or at least match the performance of, a C-class classifier with one single input feature subset. For each class, we select a desirable feature subset, which leads to the lowest classification error rate for this class using a classifier for a given feature subset search algorithm. To fairly compare all models, we propose a weight method for the class-dependent classifier, i.e., assigning a weight to each model's output before the comparison is carried out. The method's performance is evaluated on two artificial data sets and several real-world benchmark data sets, with the support vector machine (SVM) as the classifier , and with the RELIEF, class separability, and minimal-redundancy-maximal-relevancy (mRMR) as attribute importance measures. Our results indicate that the class-dependent feature subsets found by our approach can effectively remove irrelevant or redundant features, while maintaining or improving (sometimes substantially ) the classification accuracy, in comparison with other feature selection methods.

We propose a novel approach, i.e., a noisy chaotic neural network with variable thresholds (NCNN-VT), to solve the frequency assignment problem in satellite communications. The objective of this NP-complete optimization problem is to minimize cochannel interference between two satellite systems by rearranging frequency assignments. The NCNN-VT model consists N times M of noisy chaotic neurons for an N-carrier M-segment problem. The NCNN-VT facilitates the interference minimization by mapping the objective to variable thresholds (biases) of the neurons. The performance of the NCNN-VT is demonstrated by solving a set of benchmark problems and randomly generated test instances. The NCNN-VT achieves better solutions, i.e., smaller interference with much lower computation cost compared to existing algorithms.

The chaotic cellular neural network (C-CNN) has complex dynamics, including chaos, oscillations, and stable fixed points. Chaotic dynamics can help the network avoid local minima and reach the global optimum. Hence chaos can improve the performance of cellular neural networks (CNNs) on problems that have local minima in energy (cost) functions. We investigate the effect of variable thresholds in the C-CNN. We show that this threshold cannot be too large if one wishes to produce chaotic dynamics in the C-CNN, which is important to studies of chaotic communication and combinatorial optimization problem. We are particularly interested in variable thresholds because Shi and Wang (2005) showed that the objectives of the frequency assignment problem (FAP) can be mapped into thresholds of the neural network, which resulted in superior performance compared to traditional penalty approaches.

BACKGROUND: Since the single nucleotide polymorphisms (SNPs) are genetic variations which determine the difference between any two unrelated individuals, the SNPs can be used to identify the correct source population of an individual. For efficient population identification with the HapMap genotype data, as few informative SNPs as possible are required from the original 4 million SNPs. Recently, Park et al. (2006) adopted the nearest shrunken centroid method to classify the three populations, i.e., Utah residents with ancestry from Northern and Western Europe (CEU), Yoruba in Ibadan, Nigeria in West Africa (YRI), and Han Chinese in Beijing together with Japanese in Tokyo (CHB+JPT), from which 100,736 SNPs were obtained and the top 82 SNPs could completely classify the three populations. RESULTS: In this paper, we propose to first rank each feature (SNP) using a ranking measure, i.e., a modified t-test or F-statistics. Then from the ranking list, we form different feature subsets by sequentially choosing different numbers of features (e.g., 1, 2, 3, ..., 100.) with top ranking values, train and test them by a classifier, e.g., the support vector machine (SVM), thereby finding one subset which has the highest classification accuracy. Compared to the classification method of Park et al., we obtain a better result, i.e., good classification of the 3 populations using on average 64 SNPs. CONCLUSION: Experimental results show that the both of the modified t-test and F-statistics method are very effective in ranking SNPs about their classification capabilities. Combined with the SVM classifier, a desirable feature subset (with the minimum size and most informativeness) can be quickly found in the greedy manner after ranking all SNPs. Our method is able to identify a very small number of important SNPs that can determine the populations of individuals.

We aim at finding the smallest set of genes that can ensure highly accurate classification of cancers from microarray data by using supervised machine learning algorithms. The significance of finding the minimum gene subsets is three-fold: 1) It greatly reduces the computational burden and "noise" arising from irrelevant genes. In the examples studied in this paper, finding the minimum gene subsets even allows for extraction of simple diagnostic rules which lead to accurate diagnosis without the need for any classifiers. 2) It simplifies gene expression tests to include only a very small number of genes rather than thousands of genes, which can bring down the cost for cancer testing significantly. 3) It calls for further investigation into the possible biological relationship between these small numbers of genes and cancer development and treatment. Our simple yet very effective method involves two steps. In the first step, we choose some important genes using a feature importance ranking scheme. In the second step, we test the classification capability of all simple combinations of those important genes by using a good classifier. For three "small" and "simple" data sets with two, three, and four cancer (sub)types, our approach obtained very high accuracy with only two or three genes. For a "large" and "complex" data set with 14 cancer types, we divided the whole problem into a group of binary classification problems and applied the 2-step approach to each of these binary classification problems. Through this "divide-and-conquer" approach, we obtained accuracy comparable to previously reported results but with only 28 genes rather than 16,063 genes. In general, our method can significantly reduce the number of genes required for highly reliable diagnosis.

Accurate classification of cancers based on microarray gene expressions is very important for doctors to choose a proper treatment. In this paper, we apply a novel radial basis function (RBF) neural network that allows for large overlaps among the hidden kernels of the same class to this problem. We tested our RBF network in three data sets, i.e., the lymphoma data set, the small round blue cell tumors (SRBCT) data set, and the ovarian cancer data set. The results in all the three data sets show that our RBF network is able to achieve 100% accuracy with much fewer genes than the previously published methods did.

We propose an algorithm of searching for good discriminative gene sets (DGSs) in microarray cancer data, which we call active mining discriminative gene sets (AM-DGS). Tests in the leukemia data set and the prostate data set indicate that our method is able to achieve better accuracy with much smaller DGSs compared to 3 widely used methods, i.e.,TS,FS, and SVM-RFE.

In this paper we propose a novel support vector machine (SVM) with class-dependent features. According to an importance measure, e.g., the RELIEF weight measure or class separability measure, we rank the features importance for each class against the rest of classes. For each class we select an optimal feature subset using a classifier, e.g., the support vector machine (SVM). For the classification on these class-dependent feature subsets, we propose to construct a novel SVM using "one-against-all" in 2 processes: (1) construct one model for each class by training the classifier with the class's optimal feature subset; (2) during testing, each test pattern is tested on all models and the model with the maximum output decides the class of the test pattern. The method's performance is evaluated on two benchmark datasets. Our results indicate that our novel SVM classifier can effectively realize the classification of class-dependent feature subsets found by our wrapper approach which can remove irrelevant features for each class and at the same time maintain or even improve the classification accuracy in comparison with other feature selection methods.

In this paper, we propose a gradual noisy chaotic neural network (G-NCNN) to solve the NP-complete broadcast scheduling problem (BSP) in packet radio networks. The objective of the BSP is to design an optimal time-division multiple-access (TDMA) frame structure with minimal TDMA frame length and maximal channel utilization. A two-phase optimization is adopted to achieve the two objectives with two different energy functions, so that the G-NCNN not only finds the minimum TDMA frame length but also maximizes the total node transmissions. In the first phase, we propose a G-NCNN which combines the noisy chaotic neural network (NCNN) and the gradual expansion scheme to find a minimal TDMA frame length. In the second phase, the NCNN is used to find maximal node transmissions in the TDMA frame obtained in the first phase. The performance is evaluated through several benchmark examples and 600 randomly generated instances. The results show that the G-NCNN outperforms previous approaches, such as mean field annealing, a hybrid Hopfield network-genetic algorithm, the sequential vertex coloring algorithm, and the gradual neural network.

A genetic algorithm approach for segmented channel routing in field programmable gate arrays (FPGA's) is presented in this paper. The FPGA segmented channel routing problem (FSCRP) is formulated as a special case of a matrix row matching problem which is known to be NP-complete. The goal of FSCRPS is to find a conflict-free net assignment in the tracks within the channel with the minimum routing cost. Simulations on 30 benchmark instances show that GA is able to obtain better solutions compared to the gradual neural network (GNN) approach.

In this paper we propose a hybrid method to solve the broadcast scheduling problem in packet radio networks. In the first stage, we use a backtracking sequential coloring algorithm to obtain a minimal TDMA frame length and the corresponding transmission assignments. In the second stage, we employ the noisy chaotic neural network to find the maximum node transmission based on the results obtained in the previous stage. Simulation results show that this hybrid method outperforms previous approaches, such as mean field annealing, a hybrid of the Hopfield neural network and genetic algorithms, the sequential vertex coloring algorithm, and the gradual neural network.

Microarray gene expression data usually have a large number of dimensions, e.g., over ten thousand genes, and a small number of samples, e.g., a few tens of patients. In this paper, we use the support vector machine (SVM) for cancer classification with microarray data. Dimensionality reduction methods, such as principal components analysis (PCA), class-separability measure, Fisher ratio, and t-test, are used for gene selection. A voting scheme is then employed to do multi-group classification by k(k - 1) binary SVMs. We are able to obtain the same classification accuracy but with much fewer features compared to other published results.

This paper presents a new approach to address image denoising based on a new neural network, called noisy chaotic neural network (NCNN). The original Bayesian framework of image denoising is reformulated into a constrained optimization problem using continuous relaxation labeling. The NCNN, which combines the simulated annealing technique with the Hopfield neural network (HNN), is employed to solve the optimization problem. It effectively overcomes the local minima problem which may be incurred by the HNN. The experimental results show that the NCNN could offer good quality solutions.

Selection of significant genes via expression patterns is an important problem in microarray data processing. In this article, we propose and study a new method for selecting relevant genes obtained by spectral biclustering and based on similarity between genes and eigenvectors. The proposed algorithm can select a much smaller gene subset to make accurate predictions. The unsupervised gene selection method suggested in This work is demonstrated on two microarray cancer data sets, i.e., the lymphoma and the liver cancer data sets. In both examples, our method is able to identify two-gene combinations which can lead to prediction with very high accuracy.

We summarize some of the key research areas in brain sciences and neural networks that have recently been or are being worked on by researchers in Singapore. Researchers in Singapore are developing theory of neural networks, notably improved radial basis function networks, fuzzy neural networks, and fast learning neural networks. Applications of neural networks include bioinformatics, multimedia, data mining, and communications. Researchers are also working with neurophysiologists on functional brain imaging and brain disease analysis.

Recently Chen and Aihara have demonstrated both experimentally and mathematically that their chaotic simulated annealing (CSA) has better search ability for solving combinatorial optimization problems compared to both the Hopfield-Tank approach and stochastic simulated annealing (SSA). However, CSA may not find a globally optimal solution no matter how slowly annealing is carried out, because the chaotic dynamics are completely deterministic. In contrast, SSA tends to settle down to a global optimum if the temperature is reduced sufficiently slowly. Here we combine the best features of both SSA and CSA, thereby proposing a new approach for solving optimization problems, i.e., stochastic chaotic simulated annealing, by using a noisy chaotic neural network. We show the effectiveness of this new approach with two difficult combinatorial optimization problems, i.e., a traveling salesman problem and a channel assignment problem for cellular mobile communications.

Cancer classification is an important problem both for clinical treatment and for biomedical research. Considering the good performance of support vector machines (SVMs) on solving pattern recognition problems, we use a C-SVM to process the B-cell lymphoma data. The principal components analysis (PCA) is used for gene selection. A voting scheme is used to do multi-group classification by k(k-1) binary SVMs. The classification results show that SVMs are effective tools for this problem.

Both the stochastic chaotic simulated annealing and the deterministic chaotic simulated annealing are used to restore gray level images degraded by a known shift-invariant blur function and additive noise. The neural networks are modeled to represent the image whose gray level function is the simple sum of the neuron state variables. The restoration consists of two stages: parameter estimation and image reconstruction. During the first stage, parameters are estimated by comparing the energy function of the neural network to a constraint error function. The neural networks are then updated. Experiments show that noisy chaotic neural network could get good results in relatively shorter time compared to Hopfield neural network and better results compared to transiently chaotic neural network.

The channel assignment problem (CAP) in cellular systems has the task of planning the reuse of the limited available frequencies in a spectrum-efficient and interference-minimal way. We adopt an integer programming representation of the static channel assignment (SCA) formulated by Smith and Palaniswami. The tabu search (TS) algorithm is implemented to solve this problem. Ten benchmark problems are tested, and we show that the TS approach significantly outperforms other optimization techniques.

For high dimensional data, if no preprocessing is carried out before inputting patterns to classifiers, the computation required may be too heavy. For example, the number of hidden units of a radial basis function (RBF) neural network can be too large. This is not suitable for some practical applications due to speed and memory constraints. In many cases, some attributes are not relevant to concepts in the data at all. In this paper, we propose a novel separability-correlation measure (SCM) to rank the importance of attributes. According to the attribute ranking results, different attribute subsets are used as inputs to a classifier, such as an RBF neural network. Those attributes that increase the validation error are deemed irrelevant and are deleted. The complexity of the classifier can thus be reduced and its classification performance improved. Computer simulations show that our method for attribute importance ranking leads to smaller attribute subsets with higher accuracies compared with the existing SUD and Relief-F methods. We also propose a modified method for efficient construction of an RBF classifier. In this method we allow for large overlaps between clusters corresponding to the same class label. Our approach significantly reduces the structural complexity of the RBF network and improves the classification performance.

High dimensionality of data sets is a curse to classifiers. We propose to construct a novel radial basis function (RBF) classifier using class-dependent features by genetic algorithms (GA). Since each feature may have different capabilities in discriminating different classes, features should be masked differently for different classes. In our novel RBF classifier, each Gaussian kernel function of the RBF neural network is active for only a subset of patterns which are approximately of the same class. A group of Gaussian kernel functions is generated for each class. In our method, different feature masks are used for different groups of Gaussian kernel functions corresponding to different classes. The feature masks are adjusted by GA. The classification accuracy of the RBF neural network is used as the fitness function. Thus, the dimensionality of a data set is reduced. Simulations show that, with irrelevant features removed for each class, our method can lead to significant improvements on classification accuracy.

Rule extraction is a technique for knowledge discovery. Compact rules with high accuracy are desirable. Due to the curse of irrelevant features to classifiers, feature selection techniques are discussed widely. We propose to extract rules based on class-dependent features from a radial basis function (RBF) classifier by genetic algorithms (GA). Each Gaussian kernel function of the RBF neural network is active for only a subset of patterns which are approximately of the same class. Since each feature may have different capabilities in discriminating different classes, features should be masked differently for different classes. In our method, different feature masks are used for different groups of Gaussian kernel functions corresponding to different classes. The feature masks are adjusted by GA. The classification accuracy of the RBF neural network is used as the fitness function. Thus, the dimensionality of a data set is reduced. Concise rules with high accuracy are subsequently obtained based on the class-dependent features. We demonstrate our approach using computer simulations.

As an important task of data mining, extracting rules to represent the concept of numerical data is attracting much attention. We propose a novel algorithm to extract rules using genetic algorithms (GA) and the radial basis function (RBF) neural network classifier. The interval for each input in the condition part of each rule is adjusted using GA. The fitness of a chromosome is determined by the accuracy of extracted rules. The decision boundary of rules extracted is hyper-rectangular. During the training of the RBF neural network, large overlaps between clusters corresponding to the same class is allowed in order to decrease the number of hidden units while maintaining classification accuracy. The weights connecting the hidden units with the output units are then pruned. Our simulations demonstrate that our approach leads to more accurate and concise rules.

Chaotic simulated annealing (CSA) has recently been proposed and successfully used in solving combinatorial optimization problems by Chen and Aihara. In comparison with the Hopfield-Tank approach. CSA significantly improves the network's ability to find solutions of good quality and even global minima. However, CSA still uses a penalty term to enforce solution validity like the Hopfield-Tank approach. There exists a conflict between solution quality and solution validity in the penalty approach. In addition, the relative magnitude of the penalty term often needs to be determined by trial-and-error. In this paper we incorporate augmented Lagrange multipliers into CSA, obtaining a method that we call augmented Lagrange chaotic simulated annealing (AL-CSA), which eliminates the need of the penalty term and guarantees solution validity, and at the same time maintains CSA's solution quality. We demonstrate this method with the 10-city Traveling Salesman Problem.

Chaotic simulated annealing (CSA) recently proposed by Chen and Aihara (1994) has been shown to have higher searching ability for solving combinatorial optimization problems compared to both the Hopfield-Tank approach and stochastic simulated annealing (SSA). However, CSA is not guaranteed to relax to a globally optimal solution no matter how slowly annealing takes place. In contrast, SSA is guaranteed to settle down to a global minimum with probability 1 if the temperature is reduced sufficiently slowly. In this paper, we attempt to combine the best of both worlds by proposing a new approach to simulated annealing using a noisy chaotic neural network, i.e., stochastic chaotic simulated annealing (SCSA). We demonstrate this approach with the 48-city traveling salesman problem.

In hearing perception theory, spectral structure is a most important feature for speech perception, this spectral structure is not easy to be masked in noisy condition. So if this structure is extracted and enhanced, the representation will be much more robust. In this paper, we propose a new statistical dominant subspace analysis method for auditory spectrum based on SVD(Singular Values Decomposition) and signal subspace analysis method. The auditory spectrum can be decomposed into two subspaces, one is a dominant subspace, which is expanded by useful speech auditory spectrum , another subspace is sub-dominant subspace, which there is only noise information. So we analysis the auditory spectrum in the dominant subspace, the SNR will be increased. Thus this representation is much more robust. 1. COMPUTATIONAL AUDITORY MODEL AND AUDITORY SPECTRUM Speech stimulation can be represented by auditory neural system in many stages. First, it can be decomposed into many frequency bands by basilar membrane, then after processed by inner hair cell and neural fibers, it's intensity is represented by neural firing rate. This neural impulse can be transformed to auditory central system, where it can be perceived by auditory cortex[1]. In this paper, all the processing parts are integrated using digital signal processing method, when speech signal is processed by this model, auditory feature can be gotten. The basic processing frame is as in Fig.1: the system is made up of six parts, that is, the high pass filtering of outer ear and middle ear, the band pass filtering of basilar membrane, nonlinear compression and half wave rectifying of inner hair cell, low pass filtering of neural fiber, energy detection of central system, Figure1 Auditory model for speech signal processing A mathematical model is designed to simulate this auditory function, as in Fig.2, a low pass filter is used to simulate the long temporal integration mechanism. The function of outer/middle ear can be simulated by high pass filter; band pass filters for basilar membrane; halfwave rectify for inner hair cell; low pass filter for neural fiber; energy detector and log compression for neural central, at last a DCT is used to get the feature vector. Figure2 The mathematical model for Figure 1 In these modules, short term adaptation and rapid adaptation of inner hair cell and neural fiber are not considered. Also, functions of temporal integration of neural central system and low pass filtering of neural fiber are integrated as a low pass filter. Energy detector is used for the intensity detection for each frequency channel. After processed by this model, a auditory feature is gotten. The feature can be used for training and testing. In this paper, we only focus on the auditory spectrum analysis, so the auditory spectrum can be got from the energy detector of Figure 2. Visual representations of FFT, LPC, and Auditory Spectrum are drawn for comparison in figure 3.( the spectrum of a Chinese sentence ). Figure3 Top is FFT spectrum, middle is LPC spectrum. Bottom is Auditory Spectrum(AS) . From Fig.3, it is clear that AS(Auditory Spectrum) is wide band spectrum, FFT is narrow band spectrum. AS spectrum can be regarded as a smoothed spectrum of FFT spectrum in hearing perception scale(in frequency domain). It is very clear that , speech representation by auditory system is a series of time-frequency patches, these patches are different from noise patch. We can regard the speech feature as a continuous time-frequency patch with regular structure. Noise patches is random and no-regular, so we hope subspace decomposition method can help use to separate noise and speech by this property. 2 SIGNAL SUBSPACE AND SVD Signal subspace analysis method is widely used in digital signal processing and pattern recognition[2]. It is supposed that the useful information is only related with some lower dimensional subspaces, but noise is uniformly distributed in the whole measurement space (the whole Euclidean space). The subspace analysis method can decompose the whole measurement space into some useful subspaces, such as signal subspace and noise subspace or dominant subspace and subordinate subspace, then when the original feature is projected into the dominant subspaces, the dominant structure will be retained only, that is to say the subordinate feature(which including noise structure) will be reduced . From transformation view, we hope to find a new transform basis, the new feature gotten by this transformation can possess certain property, such as, each dimension of the feature vector is un-correlated or independent, etc. In this paper, we propose the subspace analysis method for the processing purpose. Singular Value Decomposition (SVD) is a very useful method for matrix structure decomposition. we give some useful formulas here which can be used later. Suppose the data matrix is n m R X × ∈ , there exist orthogonal matrices: m m m R u u U × ∈ = ) ,..., ( 1 n n m R v v V × ∈ = ) ,..., ( 1 satisfying:

Autoassociations of spatio-temporal sequences have been discussed by a number of authors. We propose a mechanism for storing and retrieving pairs of spatio-temporal sequences with the network architecture of the standard bidirectional associative memory (BAM), thereby achieving hetero-associations of spatio-temporal sequences.

In today's information age, data mining, i.e., extracting useful patterns or relationships from vast amount of data, has became increasingly important. Decision trees are currently the most popular tools for data mining. Despite many advantages in this approach, same aspects require improvements. A notable problem is known as the problem of small disjuncts, where the induced rules that cover a small amount of training cases often have high error rates. The purpose of the present paper is to show that a dynamically constructed recurrent fuzzy neural network can deal effectively with this problem.

Research in neural control mostly concentrates on indirect control schemes while insufficient attention has been paid to direct model reference adaptive control (MRAC) scheme. In addition, at present the emphasis of neural control is on parameter tuning instead of structural tuning, i.e., to find the minimal controller capable of achieving an optimal performance. The stability of the neural control schemes (i.e. the requirement of persistency of excitation and bounded learning rates) also requires more attention. Furthermore, localized architectures are needed in order to deal with the moving target problem (i.e. the difficulty for global neural networks to perform several separate computational tasks in closed-loop control). The purpose of the present paper is to show that direct MRAC using dynamically constructed neural controllers, such as the fuzzy neural and the cascade correlation, satisfy above requirements and offers a method for automatic discovery of an efficient controller.

Based on the previous work of a number of authors, we discuss an important class of neural networks which we call multi-associative neural networks (MANNs) and which associate one pattern with multiple patterns. As a computationally efficient example of such networks, we describe a specific MANN, that is, a multi-associative, dynamically generated variant of the counterpropagation network (MCPN). As an application of MANNs, we design a general system that can learn and retrieve complex spatio-temporal sequences with any MANN. This system consists of comparator units, a parallel array of MANNs, and delayed feedback lines from the output of the system to the neural network layer. During learning, pairs of sequences of spatial patterns are presented to the system and the system learns-to associate patterns at successive times in sequence. During retrieving, a cue sequence, which may be obscured by spatial noise and temporal gaps, causes the system to output the stored spatio-temporal sequence. We prove analytically that this system is capable of learning and generating any spatio-temporal sequences within the maximum complexity determined by the number of embedded MANNs, with the maximum length and number of sequences determined by the memory capacity of the embedded MANNs. To demonstrate the applicability of this general system, we present an implementation using the MCPN. The system shows desirable properties such as fast and accurate learning and retrieving, and ability to store a large number of complex sequences consisting of nonorthogonal spatial patterns.

We show that the memory capacity of the fully connected binary Hopfield network is significantly reduced by a small amount of noise in training patterns. Our analytical results obtained with the mean field method are supported by extensive computer simulations.

Chen and Aihara recently proposed a chaotic simulated annealing approach to solving optimization problems. By adding a negative self-coupling to a network model proposed earlier by Aihara et al. and gradually removing this negative self-coupling, they used the transient chaos for searching and self-organizing, thereby achieving remarkable improvement over other neural-network approaches to optimization problems with or without simulated annealing. In this paper we suggest a new approach to chaotic simulated annealing with guaranteed convergence and minimization of the energy function by gradually reducing the time step in the Euler approximation of the differential equations that describe the continuous Hopfield neural network. This approach eliminates the need to carefully select other system parameters. We also generalize the convergence theorems of Chen and Aihara to arbitrarily increasing neuronal input-output functions and to less restrictive and yet more compact forms.

We present convergence theorems for neural networks with arbitrary energy functions and discrete-time dynamics for both discrete and continuous neuronal input-output-functions. We discuss systematically how the neuronal updating rule should be extracted once an energy function is constructed for a given application, in order to guarantee the descent and minimization of the energy function as the network updates. We explain why the existing theory may lead to inaccurate results and oscillatory behaviors in the convergence process. We also point out the reason for and the side effects of using hysteresis neurons to suppress these oscillatory behaviors.

We derive learning rates such that all training patterns are equally important statistically and the learning outcome is independent of the order in which training patterns are presented, if the competitive neurons win the same sets of training patterns regardless the order of presentation. We show that under these schemes, the learning rules in the two different weight normalization approaches, the length-constraint and the sum-constraint, yield practically the same results, if the competitive neurons win the same sets of training patterns with both constraints. These theoretical results are illustrated with computer simulations.

In this paper, we show that noise injection into inputs in unsupervised learning neural networks does not improve their performance as it does in supervised learning neural networks. Specifically, we show that training noise degrades the classification ability of a sparsely connected version of the Hopfield neural network, whereas the performance of a sparsely connected winner-take-all neural network does not depend on the injected training noise.

This paper studies the effects of a time-dependent operating environment on the dynamics of a neural network. In the previous paper Wang et al. (1990) studied an exactly solvable model of a higher order neural network. We identified a bifurcation parameter for the system, i.e., the rescaled noise level, which represents the combined effects of incomplete connectivity, interference among stored patterns, and additional stochastic noise. When this bifurcation parameter assumes different but static (time-independent) values, the network shows a spectrum of dynamics ranging from fixed points, to oscillations, to chaos. This paper shows that varying operating conditions described by the time-dependence of the rescaled noise level give rise to many more interesting dynamical behaviours, such as disappearances of fixed points and transitions between periodic oscillations and deterministic chaos. These results suggest that a varying environment, such as the one studied in the present model, may be used to facilitate memory retrieval if dynamic states are used for information storage in a neural network.