Search

Software defect prediction is one of the most active research areas in software engineering and plays an important role in software quality assurance. In recent years, many new defect prediction studies have been proposed. There are four main aspects of research: machine learning-based prediction algorithms, manipulating the data, effor-softaware prediction and empirical studies. The research community is still facing many challenges in constructing methods, and there are also many research opportunities in the meantime. This paper proposes a method of applying spiking neural network to software defect prediction. The software defect prediction model is constructed by feed-forward spiking neural networks and trained by spike train learning algorithm. This model uses the existing project data sets to predict software defects projects. Extensive experiments on 28 public projects from five data sources indicate that the effectiveness of the proposed approach with respect to the considered metrics.

In biological nervous systems, the synaptic transmission of information is a complex process through the release of neurotransmitters. The input of multiple signals shows nonlinear interaction characteristics in synapses, so nonlinear synaptic interaction is considered as an important part of biological neural networks. At present, most artificial neural networks simplify synapses into a linear structure. Considering the nonlinear interaction of the input multiple signals of synapse, this paper proposes an online supervised learning algorithm for spiking neural networks based on nonlinear synaptic kernels, which can implement the complex spatio-temporal pattern learning of spike trains. The algorithm is successfully applied to learn sequences of spikes. In addition, different learning parameters are analyzed, such as synaptic kernel. The experimental results show that the proposed algorithm has high learning accuracy.

Deep spiking belief network (DSBN) uses unsupervised layer-wise pre-training method to train the network weights, it is stacked with the spike neural machine (SNM) modules. However, the synaptic weights of SNMs are difficult to pre-training through simple and effective approach for spike-train driven networks. This paper proposes a new algorithm that uses unsupervised multi-spike learning rule to train SNMs, which can implement the complex spatio-temporal pattern learning of spike trains. The spike signals first propagate in the forward direction, and then are reconstructed in the reverse direction, and the synaptic weights are adjusted according to the reconstruction error. The algorithm is successfully applied to spike train patterns, the module parameters are analyzed, such as the neuron number and learning rate in the SNMs. In addition, the low reconstruction errors of DSBNs are shown by the experimental results.

Short texts have become a kind of prevalent source of information, and the classification of these short texts in various forms is valuable to many applications. However, most existing short text classification approaches circumvent the sparsity problem by extending short texts (or their feature representations) or exploiting additional information to adapt short texts to traditional text classification approaches. In this paper, we try to solve the sparsity problem of short text classification in a different direction: adapting the classifier to short texts. We propose a sparse representation short text classification method based on entropy weighted constraint. The main idea behind this study is to consider that the short texts are similar in potentially specific subspace. Specifically, we first introduce word embedding to represent the initial sparse representation dictionary, and then a fast feature subset selection algorithm is used to filter the dictionary. Again, we design an objective function based on sparse representation of entropy weight constraint. The optimal value of the objective function is obtained by Lagrange multiplier method. Finally, the distance between the short text to be classified and the short text in each class is calculated under the subspace, and the short text is classified according to three classification rules. Experiments over five datasets show that the proposed approach can effectively alleviate the problem of sparse feature of short text and is more efficient and effective than the existing short text classification method.

Measuring the similarity between short texts is made difficult by the fact that two texts that are semantically related may not contain any words in common. In this paper, we propose a novel short text similarity measure which aggregates coupled semantic relation (CSR) and strong classification features (SCF) to provide a richer semantic context. On the one hand, CSR considers both intra-relation (i.e. co-occurrence of terms based on the modified weighting strategy) and inter-relation (i.e. dependency of terms via paths that connect linking terms) between a pair of terms. On the other hand, Based on SCF for similarity measure is established based on the idea that the more similar two texts are, the more features of strong classification they share. Finally, we combine the above two techniques to address the semantic sparseness of short text. We carry out extensive experiments on real world short texts. The results demonstrate that our method significantly outperforms baseline methods on several evaluation metrics.

The liquid state machines have been well applied for solving large-scale spatio-temporal pattern recognition problems. The current supervised learning algorithms for the liquid state machines of spiking neurons generally only adjust the synaptic weights in the output layer, the synaptic weights of input and hidden layers are generated in the process of network structure initialization and no longer change. That is to say, the hidden layer is a static network, which usually neglects the dynamic characteristics of the liquid state machines. Therefore, a new supervised learning algorithm for the liquid state machines of spiking neurons based on bidirectional modification is proposed, which not only adjusts the synaptic weights in the output layer, but also changes the synaptic weights in the input and hidden layers. The algorithm is successfully applied to the spike trains learning. The experimental results show that the proposed learning algorithm can effectively learn the spike trains pattern with different learning parameter.

Clustering is one of the most basic unsupervised learning problems in the field of machine learning and its main goal is to separate data into clusters with similar data points. Because of various redundant and complex structures for the raw data, the general algorithm usually is difficult to separate different clusters from the data and the effect is not obvious. Deep learning is a technology that automatically learns nonlinear and more conducive clustering features from complex data structures. This paper presents a deep clustering algorithm based on self-organizing map neural network. This method combines the feature learning ability of stacked auto-encoder from the raw data and feature clustering with unsupervised learning of self-organizing map neural network. It is aim to achieve the greatest separability for the data space. Through the experimental analysis and comparison, the proposed algorithm has better recognition rate, and improves the clustering performance on low and high dimension data.

The neuron classification problem is significant for understanding structure-function relationships in computational neuroscience. Advances in recent years have accelerated the speed of data collection, resulting in a large amount of data on the geometric, morphological, physiological, and molecular characteristics of neurons. These data encourage researchers to strive for automated neuron classification through powerful machine learning techniques. This paper extracts a statistical dataset of 43 geometrical features obtained from 116 human neurons, and proposes a neuronal morphology classification approach based on deep residual neural networks with feature scaling. The approach is applied to classify 18 types of human neurons and compares the accuracy of different number of residual block. Then, we also compare the accuracy between the proposed approach and other mainstream machine learning approaches, the classification accuracy of our approach is 100% in the training set and the testing set accuracy is 76.96%. The experimental results show that the deep residual neural network model has better classification accuracy for human neurons.

The recurrent spiking neural networks include complex structures and implicit nonlinear mechanisms, the formulation of efficient supervised learning algorithm is difficult and remains an important problem in the research area. This paper proposes a new supervised multi-spike learning algorithm for recurrent spiking neural networks, which can implement the complex spatiotemporal pattern learning of spike trains. Using information encoded in precisely timed spike trains and their inner product operators, the error function is firstly constructed. Furthermore, the proposed algorithm defines the learning rules of synaptic weights based on inner product of spike trains. The algorithm is successfully applied to learn spike train patterns, and the high learning accuracy and efficiency are shown by the experimental results. In addition, the network structure parameters are analyzed, such as the neuron number and connectivity degree in the recurrent layer of spiking neural networks.

The complex network structure of biological brains is obtained through the developmental processes. Type and complexity of network structure directly reflect the ability of the network to deal with information processing. In this paper, we propose a developmental method for creating recurrent spiking neural networks based on genetic regulatory network model. This research investigates the developmental process of spiking neural networks, and analyzes the network structure in the different parameter settings, such as the number of regulatory nodes, the weights scale of genetic networks, and the developmental scale. The experimental results show that the developmental spiking neural networks have the similar topological characteristics as biological networks, namely scale-free and small-world properties.

Inspired by the evolution of biological brains, the study of neurally-driven evolved autonomous agents has received more and more attention. In this paper, we propose an evolutionary algorithm for neurally-driven autonomous agents, each agent is controlled by a spiking neural network, and the network receives the sensory inputs and processes the motor outputs through the encoded spike information. The controlling spiking neural networks of autonomous agents are developed by the evolutionary algorithms that apply some of genetic operators and selection to a population of agents that undergo evolution. The corresponding food gathering experiment results show that the autonomous agents appear intelligent behaviours for the simulation environment. Additionally, the parameters of networks and agents play an important role in the evolutionary process.

The purpose of supervised learning with temporal encoding for spiking neurons is to make the neurons emit arbitrary spike trains in response to given synaptic inputs. Recent years, the supervised learning algorithms based on synaptic plasticity have developed rapidly. As one of the most efficient supervised learning algorithms, the remote supervised method (ReSuMe) uses the conventional pair-based spike-timing-dependent plasticity rule, which depends on the precise timing of presynaptic and postsynaptic spikes. In this paper, using the triplet-based spike-timing-dependent plasticity, which is a powerful synaptic plasticity rule and acts beyond the classical rule, a novel supervised learning algorithm, dubbed T-ReSuMe, is presented to improve ReSuMe’s performance. The proposed algorithm is successfully applied to various spike trains learning tasks, in which the desired spike trains are encoded by Poisson process. The experimental results show that T-ReSuMe has higher learning accuracy and fewer iteration epoches than the traditional ReSuMe, so it is effective for solving complex spatio-temporal pattern learning problems.

Spiking neural networks are shown to be suitable tools for the processing of spatio-temporal information. However, due to their intricately discontinuous and implicit nonlinear mechanisms, the formulation of efficient supervised learning algorithms for spiking neural networks is difficult, which has become an important problem in the research area. This paper presents a new supervised, multi-spike learning algorithm for spiking neurons, which can implement the complex spatio-temporal pattern learning of spike trains. The proposed algorithm firstly defines nonlinear inner products operators to mathematically describe and manipulate spike trains, and then derive the learning rule from the common Widrow-Hoff rule with the nonlinear inner products of spike trains. The algorithm is successfully applied to learn sequences of spikes. The experimental results show that the proposed algorithm is effective for solving complex spatio-temporal pattern learning problems.

In this paper, we present a microblog recommendation algorithm based on multi-tag correlation. Firstly, a tag retrieval strategy is designed to add tags for unlabeled users, the initial user-tag matrix is then constructed and user-tag weights are set. In order to represent user interests accurately, we fully investigate the associations between the tags. Both inner and outer correlation between tags are defined to conquer the problem of sparsity of user-tag matrix. The user interests can then be decided and microblogs can be recommended to users. Experimental results show that the algorithm is effective for microblog recommendation.

The online learning algorithm is shown to be more appropriate and effective for the processing of spatiotemporal information, but very little researches have been achieved in developing online learning approaches for spiking neural networks. This paper presents an online supervised learning algorithm based on nonlinear spike train kernels to process the spatiotemporal information, which is more biological interpretability. The main idea adopts online learning algorithm and selects a suitable kernel function. At first, the Laplacian kernel function is selected, however, in some ways, the spike trains expressed by the simple kernel function are linear in the postsynaptic neuron. Then this paper uses nonlinear functions to transform the spike train model and presents the detail experimental analysis. The proposed learning algorithm is evaluated by the learning of spike trains, and the experimental results show that the online nonlinear spike train kernels own a super-duper learning effect.

In this paper, we present a semi-supervised clustering method for microblog in which both word-level and microblog document-level constraints are automatically generated totally based on statistical information rather than any kind of external knowledge. The key idea is first to explore term correlation data, which investigates both inter and intra correlation of words, and the initial similarity between words can therefore be deduced. And then an iterative method is established to calculate both word similarity and microblog similarity. The mechanism of incorporating dual constraints is presented based on word similarity and microblog similarity. We then formulate short text clustering problem as a non-negative matrix factorization based on dual constraints. Empirical study of two real-world dataset shows the superior performance of our framework in handling noisy and microblogs.

Inspired by the structure and behavior of the human visual system, an automatic image segmentation algorithm based on a spiking neural network model is proposed. At first, the image pixel values are encoded into the timing of spikes of neurons using the time-to-first-spike coding strategy. Then the segmentation model of spiking neural networks is applied to generate the matrix of spike timing for the visual image. Finally, using the maximum Shannon entropy as the fitness function of genetic algorithm, the evolved segmentation threshold is obtained to segment the visual image. The experimental results show that the method can obtain the optimum segmentation threshold, and achieve satisfactory segmentation results for different images.