Search

Dropout and DropConnect are two techniques to facilitate the regularization of neural network models, having achieved the state-of-the-art results in several benchmarks. In this paper, to improve the generalization capability of spiking neural networks (SNNs), the two drop techniques are first applied to the state-of-the-art SpikeProp learning algorithm resulting in two improved learning algorithms called SPDO (SpikeProp with Dropout) and SPDC (SpikeProp with DropConnect). In view that a higher membrane potential of a biological neuron implies a higher probability of neural activation, three adaptive drop algorithms, SpikeProp with Adaptive Dropout (SPADO), SpikeProp with Adaptive DropConnect (SPADC), and SpikeProp with Group Adaptive Drop (SPGAD), are proposed by adaptively adjusting the keep probability for training SNNs. A convergence theorem for SPDC is proven under the assumptions of the bounded norm of connection weights and a finite number of equilibria. In addition, the five proposed algorithms are carried out in a collaborative neurodynamic optimization framework to improve the learning performance of SNNs. The experimental results on the four benchmark data sets demonstrate that the three adaptive algorithms converge faster than SpikeProp, SPDO, and SPDC, and the generalization errors of the five proposed algorithms are significantly smaller than that of SpikeProp. Furthermore, the experimental results also show that the five algorithms based on collaborative neurodynamic optimization can be improved in terms of several measures. [ABSTRACT FROM AUTHOR]

Improved data visualization will be a significant tool to enhance cluster analysis. In this paper, an information-theoretic-based method for cluster visualization using self-organizing maps (SOMs) is presented. The information-theoretic visualization (IT-vis) has the same structure as the unified distance matrix, but instead of depicting Euclidean distances between adjacent neurons, it displays the similarity between the distributions associated with adjacent neurons. Each SOM neuron has an associated subset of the data set whose cardinality controls the granularity of the IT-vis and with which the first- and second-order statistics are computed and used to estimate their probability density functions. These are used to calculate the similarity measure, based on Renyi’s quadratic cross entropy and cross information potential (CIP). The introduced visualizations combine the low computational cost and kernel estimation properties of the representative CIP and the data structure representation of a single-linkage-based grouping algorithm to generate an enhanced SOM-based visualization. The visual quality of the IT-vis is assessed by comparing it with other visualization methods for several real-world and synthetic benchmark data sets. Thus, this paper also contains a significant literature survey. The experiments demonstrate the IT-vis cluster revealing capabilities, in which cluster boundaries are sharply captured. Additionally, the information-theoretic visualizations are used to perform clustering of the SOM. Compared with other methods, IT-vis of large SOMs yielded the best results in this paper, for which the quality of the final partitions was evaluated using external validity indices. [ABSTRACT FROM AUTHOR]

In this paper, we first address the dynamics of the elegant multivalued quaternionic Hopfield neural network (MV-QHNN) proposed by Minemoto et al. Contrary to what was expected, we show that the MV-QHNN, as well as one of its variation, does not always come to rest at an equilibrium state under the usual conditions. In fact, we provide simple examples in which the network yields a periodic sequence of quaternionic state vectors. Afterward, we turn our attention to the continuous-valued quaternionic Hopfield neural network (CV-QHNN), which can be derived from the MV-QHNN by means of a limit process. The CV-QHNN can be implemented more easily than the MV-QHNN model. Furthermore, the asynchronous CV-QHNN always settles down into an equilibrium state under the usual conditions. Theoretical issues are all illustrated by examples in this paper. [ABSTRACT FROM AUTHOR]

This paper investigates the dynamical properties of a network of neurons, each of which implements an asynchronous mapping based on polynomial growth transforms. In the first part of this paper, we present a geometric approach for visualizing the dynamics of the network where each of the neurons traverses a trajectory in a dual optimization space, whereas the network itself traverses a trajectory in an equivalent primal optimization space. We show that as the network learns to solve basic classification tasks, different choices of primal-dual mapping produce unique but interpretable neural dynamics like noise shaping, spiking, and bursting. While the proposed framework is general enough, in this paper, we demonstrate its use for designing support vector machines (SVMs) that exhibit noise-shaping properties similar to those of $\Sigma \Delta $ modulators, and for designing SVMs that learn to encode information using spikes and bursts. It is demonstrated that the emergent switching, spiking, and burst dynamics produced by each neuron encodes its respective margin of separation from a classification hyperplane whose parameters are encoded by the network population dynamics. We believe that the proposed growth transform neuron model and the underlying geometric framework could serve as an important tool to connect well-established machine learning algorithms like SVMs to neuromorphic principles like spiking, bursting, population encoding, and noise shaping. [ABSTRACT FROM AUTHOR]

Memory is one of the most important mechanisms in recurrent neural networks (RNNs) learning. It plays a crucial role in practical applications, such as sequence learning. With a good memory mechanism, long term history can be fused with current information, and can thus improve RNNs learning. Developing a suitable memory mechanism is always desirable in the field of RNNs. This paper proposes a novel memory mechanism for RNNs. The main contributions of this paper are: 1) an auxiliary memory unit (AMU) is proposed, which results in a new special RNN model (AMU-RNN), separating the memory and output explicitly and 2) an efficient learning algorithm is developed by employing the technique of error flow truncation. The proposed AMU-RNN model, together with the developed learning algorithm, can learn and maintain stable memory over a long time range. This method overcomes both the learning conflict problem and gradient vanishing problem. Unlike the traditional method, which mixes the memory and output with a single neuron in a recurrent unit, the AMU provides an auxiliary memory neuron to maintain memory in particular. By separating the memory and output in a recurrent unit, the problem of learning conflicts can be eliminated easily. Moreover, by using the technique of error flow truncation, each auxiliary memory neuron ensures constant error flow during the learning process. The experiments demonstrate good performance of the proposed AMU-RNNs and the developed learning algorithm. The method exhibits quite efficient learning performance with stable convergence in the AMU-RNN learning and outperforms the state-of-the-art RNN models in sequence generation and sequence classification tasks. [ABSTRACT FROM AUTHOR]

In the past few years, the demand for real-time hardware implementations of deep neural networks (DNNs), especially convolutional neural networks (CNNs), has dramatically increased, thanks to their excellent performance on a wide range of recognition and classification tasks. When considering real-time action recognition and video/image classification systems, latency is of paramount importance. Therefore, applications strive to maximize the accuracy while keeping the latency under a given application-specific maximum: in most cases, this threshold cannot exceed a few hundred milliseconds. Until now, the research on DNNs has mainly focused on achieving a better classification or recognition accuracy, whereas very few works in literature take in account the computational complexity of the model. In this paper, we propose an efficient computational method, which is inspired by a computational core of fully connected neural networks, to process convolutional layers of state-of-the-art deep CNNs within strict latency requirements. To this end, we implemented our method customized for VGG and VGG-based networks which have shown state-of-the-art performance on different classification/recognition data sets. The implementation results in 65-nm CMOS technology show that the proposed accelerator can process convolutional layers of VGGNet up to 9.5 times faster than state-of-the-art accelerators reported to-date while occupying 3.5 mm2. [ABSTRACT FROM AUTHOR]

There is a biological evidence to prove information is coded through precise timing of spikes in the brain. However, training a population of spiking neurons in a multilayer network to fire at multiple precise times remains a challenging task. Delay learning and the effect of a delay on weight learning in a spiking neural network (SNN) have not been investigated thoroughly. This paper proposes a novel biologically plausible supervised learning algorithm for learning precisely timed multiple spikes in a multilayer SNNs. Based on the spike-timing-dependent plasticity learning rule, the proposed learning method trains an SNN through the synergy between weight and delay learning. The weights of the hidden and output neurons are adjusted in parallel. The proposed learning method captures the contribution of synaptic delays to the learning of synaptic weights. Interaction between different layers of the network is realized through biofeedback signals sent by the output neurons. The trained SNN is used for the classification of spatiotemporal input patterns. The proposed learning method also trains the spiking network not to fire spikes at undesired times which contribute to misclassification. Experimental evaluation on benchmark data sets from the UCI machine learning repository shows that the proposed method has comparable results with classical rate-based methods such as deep belief network and the autoencoder models. Moreover, the proposed method can achieve higher classification accuracies than single layer and a similar multilayer SNN. [ABSTRACT FROM AUTHOR]

In this paper, we propose an online learning algorithm for supervised learning in multilayer spiking neural networks (SNNs). It is found that the spike timings of neurons in an SNN can be exploited to estimate the gradients that are associated with each synapse. With the proposed method of estimating gradients, learning similar to the stochastic gradient descent process employed in a conventional artificial neural network (ANN) can be achieved. In addition to the conventional layer-by-layer backpropagation, a one-pass direct backpropagation is possible using the proposed learning algorithm. Two neural networks, with one and two hidden layers, are employed as examples to demonstrate the effectiveness of the proposed learning algorithms. Several techniques for more effective learning are discussed, including utilizing a random refractory period to avoid saturation of spikes, employing a quantization noise injection technique and pseudorandom initial conditions to decorrelate spike timings, in addition to leveraging the progressive precision in an SNN to reduce the inference latency and energy. Extensive parametric simulations are conducted to examine the aforementioned techniques. The learning algorithm is developed with the considerations of ease of hardware implementation and relative compatibility with the classic ANN-based learning. Therefore, the proposed algorithm not only enjoys the high energy efficiency and good scalability of an SNN in its specialized hardware but also benefits from the well-developed theory and techniques of conventional ANN-based learning. The Modified National Institute of Standards and Technology database benchmark test is conducted to verify the newly proposed learning algorithm. Classification correct rates of 97.2% and 97.8% are achieved for the one-hidden-layer and two-hidden-layer neural networks, respectively. Moreover, a brief discussion of the hardware implementations is presented for two mainstream architectures. [ABSTRACT FROM AUTHOR]

In a real brain, the act of perception is a bidirectional process, depending on both feedforward sensory pathways and feedback pathways that carry expectations. We are interested in how such a neural network might emerge from a biologically plausible learning rule. Other neural network learning methods either only apply to feedforward networks, or employ assumptions (such as weight copying) that render them unlikely in a real brain. Predictive estimators (PEs) offer a better solution to this bidirectional learning scenario. However, PEs also depend on weight copying. In this paper, we propose the symmetric PE (SPE), an architecture that can learn both feedforward and feedback connection weights individually using only locally available information. We demonstrate that the SPE can learn complicated mappings without the use of weight copying. The SPE networks also show promise in deeper architectures. [ABSTRACT FROM AUTHOR]

An extreme learning machine (ELM) can be regarded as a two-stage feed-forward neural network (FNN) learning system that randomly assigns the connections with and within hidden neurons in the first stage and tunes the connections with output neurons in the second stage. Therefore, ELM training is essentially a linear learning problem, which significantly reduces the computational burden. Numerous applications show that such a computation burden reduction does not degrade the generalization capability. It has, however, been open that whether this is true in theory. The aim of this paper is to study the theoretical feasibility of ELM by analyzing the pros and cons of ELM. In the previous part of this topic, we pointed out that via appropriately selected activation functions, ELM does not degrade the generalization capability in the sense of expectation. In this paper, we launch the study in a different direction and show that the randomness of ELM also leads to certain negative consequences. On one hand, we find that the randomness causes an additional uncertainty problem of ELM, both in approximation and learning. On the other hand, we theoretically justify that there also exist activation functions such that the corresponding ELM degrades the generalization capability. In particular, we prove that the generalization capability of ELM with Gaussian kernel is essentially worse than that of FNN with Gaussian kernel. To facilitate the use of ELM, we also provide a remedy to such a degradation. We find that the well-developed coefficient regularization technique can essentially improve the generalization capability. The obtained results reveal the essential characteristic of ELM in a certain sense and give theoretical guidance concerning how to use ELM. [ABSTRACT FROM AUTHOR]

In this paper, we consider a modified error-correction learning rule for the multilayer neural network with multivalued neurons (MLMVN). This modification is based on the soft margins technique, which leads to the minimization of the distance between a cluster center and the learning samples belonging to this cluster. MLMVN has a derivative-free learning algorithm based on the error-correction learning rule and demonstrate a higher functionality and better generalization capability than a number of other machine learning techniques. The discrete $k$ -valued multivalued neuron activation function divides a complex plane into $k$ equal sectors. For more efficient and reliable solving of classification problems it is possible to modify the MLMVN learning algorithm in such a way that learning samples belonging to different classes (clusters) will be located as close as possible to the bisector of a desired sector (the cluster center) and as far as possible from each other, respectively. Such a modification based on the soft margins learning technique is considered in this paper. This modified learning algorithm improves the generalization capability of MLMVN when solving classification problems. [ABSTRACT FROM AUTHOR]

In this paper, we first extend the Wirtinger derivative which is defined for complex functions to hyperbolic functions, and derive the hyperbolic gradient operator yielding the steepest descent direction by using it. Next, we derive the hyperbolic backpropagation learning algorithms for some multilayered hyperbolic neural networks (NNs) using the hyperbolic gradient operator. It is shown that the use of the Wirtinger derivative reduces the effort necessary for the derivation of the learning algorithms by half, simplifies the representation of the learning algorithms, and makes their computer programs easier to code. In addition, we discuss the differences between the derived Hyperbolic-BP rules and the complex-valued backpropagation learning rule (Complex-BP). Finally, we make some experiments with the derived learning algorithms. As a result, we find that the convergence rates of the Hyperbolic-BP learning algorithms are high even if the fully activation functions are used, and discover that the Hyperbolic-BP learning algorithm for the hyperbolic NN with the split-type hyperbolic activation function has an ability to learn hyperbolic rotation as its inherent property. [ABSTRACT FROM AUTHOR]

In this paper, a variant of gray wolf optimization (GWO) that uses reinforcement learning principles combined with neural networks to enhance the performance is proposed. The aim is to overcome, by reinforced learning, the common challenge of setting the right parameters for the algorithm. In GWO, a single parameter is used to control the exploration/exploitation rate, which influences the performance of the algorithm. Rather than using a global way to change this parameter for all the agents, we use reinforcement learning to set it on an individual basis. The adaptation of the exploration rate for each agent depends on the agent’s own experience and the current terrain of the search space. In order to achieve this, experience repository is built based on the neural network to map a set of agents’ states to a set of corresponding actions that specifically influence the exploration rate. The experience repository is updated by all the search agents to reflect experience and to enhance the future actions continuously. The resulted algorithm is called experienced GWO (EGWO) and its performance is assessed on solving feature selection problems and on finding optimal weights for neural networks algorithm. We use a set of performance indicators to evaluate the efficiency of the method. Results over various data sets demonstrate an advance of the EGWO over the original GWO and over other metaheuristics, such as genetic algorithms and particle swarm optimization. [ABSTRACT FROM PUBLISHER]

Deep neural networks (DNNs) have demonstrated impressive performance in complex machine learning tasks such as image classification or speech recognition. However, due to their multilayer nonlinear structure, they are not transparent, i.e., it is hard to grasp what makes them arrive at a particular classification or recognition decision, given a new unseen data sample. Recently, several approaches have been proposed enabling one to understand and interpret the reasoning embodied in a DNN for a single test image. These methods quantify the “importance” of individual pixels with respect to the classification decision and allow a visualization in terms of a heatmap in pixel/input space. While the usefulness of heatmaps can be judged subjectively by a human, an objective quality measure is missing. In this paper, we present a general methodology based on region perturbation for evaluating ordered collections of pixels such as heatmaps. We compare heatmaps computed by three different methods on the SUN397, ILSVRC2012, and MIT Places data sets. Our main result is that the recently proposed layer-wise relevance propagation algorithm qualitatively and quantitatively provides a better explanation of what made a DNN arrive at a particular classification decision than the sensitivity-based approach or the deconvolution method. We provide theoretical arguments to explain this result and discuss its practical implications. Finally, we investigate the use of heatmaps for unsupervised assessment of the neural network performance. [ABSTRACT FROM PUBLISHER]

The increasing liberalization of European electricity markets, the growing proportion of intermittent renewable energy being fed into the energy grids, and also new challenges in the patterns of energy consumption (such as electric mobility) require flexible and intelligent power grids capable of providing efficient, reliable, economical, and sustainable energy production and distribution. From the supplier side, particularly, the integration of renewable energy sources (e.g., wind and solar) into the grid imposes an engineering and economic challenge because of the limited ability to control and dispatch these energy sources due to their intermittent characteristics. Time-series prediction of wind speed for wind power production is a particularly important and challenging task, wherein prediction intervals (PIs) are preferable results of the prediction, rather than point estimates, because they provide information on the confidence in the prediction. In this paper, two different machine learning approaches to assess PIs of time-series predictions are considered and compared: 1) multilayer perceptron neural networks trained with a multiobjective genetic algorithm and 2) extreme learning machines combined with the nearest neighbors approach. The proposed approaches are applied for short-term wind speed prediction from a real data set of hourly wind speed measurements for the region of Regina in Saskatchewan, Canada. Both approaches demonstrate good prediction precision and provide complementary advantages with respect to different evaluation criteria. [ABSTRACT FROM PUBLISHER]

This article presents the design and implementation of an embedded programmable processor with a custom instruction set architecture for efficient realization of artificial neural networks (ANNs). The ANN processor architecture is scalable, supporting an arbitrary number of layers and number of artificial neurons (ANs) per layer. Moreover, the processor supports ANNs with arbitrary interconnect structures among ANs to realize both feed-forward and dynamic recurrent networks. The processor architecture is customizable in which the numerical representation of inputs, outputs, and signals among ANs can be parameterized to an arbitrary fixed-point format. An ASIC implementation of the designed programmable ANN processor for networks with up to 512 ANs and 262,000 interconnects is presented and is estimated to occupy 2.23 mm2 of silicon area and consume 1.25 mW of power from a 1.6 V supply while operating at 74 MHz in a standard 32-nm CMOS technology. In order to assess and compare the efficiency of the designed ANN processor, we have designed and implemented a dedicated reconfigurable hardware architecture for the direct realization of ANNs. Characteristics and implementation results of the designed programmable ANN processor and the dedicated ANN hardware on a Xilinx Artix-7 field-programmable gate array (FPGA) are presented and compared using two benchmarks, the MNIST benchmark using a feed-forward ANN and a movie review sentiment analysis benchmark using a recurrent neural network. [ABSTRACT FROM AUTHOR]

Recent research has shown the potential capability of spiking neural networks (SNNs) to model complex information processing in the brain. There is biological evidence to prove the use of the precise timing of spikes for information coding. However, the exact learning mechanism in which the neuron is trained to fire at precise times remains an open problem. The majority of the existing learning methods for SNNs are based on weight adjustment. However, there is also biological evidence that the synaptic delay is not constant. In this paper, a learning method for spiking neurons, called delay learning remote supervised method (DL-ReSuMe), is proposed to merge the delay shift approach and ReSuMe-based weight adjustment to enhance the learning performance. DL-ReSuMe uses more biologically plausible properties, such as delay learning, and needs less weight adjustment than ReSuMe. Simulation results have shown that the proposed DL-ReSuMe approach achieves learning accuracy and learning speed improvements compared with ReSuMe. [ABSTRACT FROM PUBLISHER]

This paper presents a fuzzy extreme learning machine (F-ELM) that embeds fuzzy membership functions and rules into the hidden layer of extreme learning machine (ELM). Similar to the concept of ELM that employed the random initialization technique, three parameters of F-ELM are randomly assigned. They are the standard deviation of the membership functions, matrix-C (rule-combination matrix), and matrix-D [don’t care (DC) matrix]. Fuzzy if-then rules are formulated by the rule-combination Matrix of F-ELM, and a DC approach is adopted to minimize the number of input attributes in the rules. Furthermore, F-ELM utilizes the output weights of the ELM to form the target class and confidence factor for each of the rules. This is to indicate that the corresponding consequent parameters are determined analytically. The operations of F-ELM are equivalent to a fuzzy inference system. Several benchmark data sets and a real world fault detection and diagnosis problem have been used to empirically evaluate the efficacy of the proposed F-ELM in handling pattern classification tasks. The results show that the accuracy rates of F-ELM are comparable (if not superior) to ELM with distinctive ability of providing explicit knowledge in the form of interpretable rule base. [ABSTRACT FROM PUBLISHER]

In this paper a multi-level fuzzy min-max neural network classifier (MLF), which is a supervised learning method, is described. MLF uses basic concepts of the fuzzy min-max (FMM) method in a multi-level structure to classify patterns. This method uses separate classifiers with smaller hyperboxes in different levels to classify the samples that are located in overlapping regions. The final output of the network is formed by combining the outputs of these classifiers. MLF is capable of learning nonlinear boundaries with a single pass through the data. According to the obtained results, the MLF method, compared to the other FMM networks, has the highest performance and the lowest sensitivity to maximum size of the hyperbox parameter (\theta), with a training accuracy of 100% in most cases. [ABSTRACT FROM AUTHOR]