Roberto Antonio Vázquez

Training spiking neural models using cuckoo search algorithm

Several meta-heuristic algorithms have been proposed in the last years for solving a wide range of optimization problems. Cuckoo Search Algorithm (CS) is a novel meta-heuristic based on the obligate brood parasitic behaviour of some cuckoo species in combination with the Levy flight behavior of some birds and fruit flies. This algorithm has been applied in a wide range of optimization problems; nonetheless, their promising results suggest its application in the field of artificial neural networks, specially during the adjustment of the synaptic weights. On the other hand, spiking neurons are neural models that try to simulate the behavior of biological neurons when they are excited with an input current (input pattern) during a certain period time. Instead of generating a response in its output every iteration, as classical neurons do, this model generates a response (spikes or spike train) only when the model reaches a specific threshold. This response could be coded into a firing rate and perform a pattern classification task according to the firing rate generated with the input current. To perform a classification task the model ought to exhibit the next behavior: patterns from the same class must generate similar firing rates and patterns from other classes have to generate firing rates sufficiently dissimilar to differentiate among the classes. The model needs of a training phase aimed to adjust their synaptic weights and exhibit the desired behavior. In this paper, we describe how the CS algorithm can be useful to train a spiking neuron to be applied in a pattern classification task. The accuracy of the methodology is tested using several pattern recognition problems.

Classification of DNA microarrays using artificial neural networks and ABC algorithm

Graphical abstractDisplay Omitted HighlightsABC algorithm discovered the best set of genes to classify correctly cancer samples.With less than 1% of information is possible to classify with an accuracy of 93.2%.Multilayer perceptron performs better than radial basis function network. DNA microarray is an efficient new technology that allows to analyze, at the same time, the expression level of millions of genes. The gene expression level indicates the synthesis of different messenger ribonucleic acid (mRNA) molecule in a cell. Using this gene expression level, it is possible to diagnose diseases, identify tumors, select the best treatment to resist illness, detect mutations among other processes. In order to achieve that purpose, several computational techniques such as pattern classification approaches can be applied. The classification problem consists in identifying different classes or groups associated with a particular disease (e.g., various types of cancer, in terms of the gene expression level). However, the enormous quantity of genes and the few samples available, make difficult the processes of learning and recognition of any classification technique. Artificial neural networks (ANN) are computational models in artificial intelligence used for classifying, predicting and approximating functions. Among the most popular ones, we could mention the multilayer perceptron (MLP), the radial basis function neural network (RBF) and support vector machine (SVM). The aim of this research is to propose a methodology for classifying DNA microarray. The proposed method performs a feature selection process based on a swarm intelligence algorithm to find a subset of genes that best describe a disease. After that, different ANN are trained using the subset of genes. Finally, four different datasets were used to validate the accuracy of the proposal and test the relevance of genes to correctly classify the samples of the disease.

Design of artificial neural networks using differential evolution algorithm

The design of an Artificial Neural Network (ANN) is a difficult task for it depends on the human experience. Moreover it needs a process of testing and error to select which kind of a transfer function and which algorithm should be used to adjusting the synaptic weights in order to solve a specific problem. In the last years, bio-inspired algorithms have shown their power in different nonlinear optimization problems. Due to their efficiency and adaptability, in this paper we explore a new methodology to automatically design an ANN based on the Differential Evolution (DE) algorithm. The proposed method is capable to find the topology, the synaptic weights and the transfer functions to solve a given pattern classification problems.

Integrate and Fire neurons and their application in pattern recognition

In this paper, it is shown how a Leaky Integrate and Fire (LIF) neuron can be applied to solve non-linear pattern recognition problems. Given a set of input patterns belonging to K classes, each input pattern is transformed into an input signal, then the LIF neuron is stimulated during T ms and finally the firing rate is computed. After adjusting the synaptic weights of the neuron model, we expect that input patterns belonging to the same class generate almost the same firing rate and input patterns belonging to different classes generate firing rates different enough to discriminate among the different classes. At last, a comparison between a feed-forward neural network and the LIF neuron is presented when applied to solve non-linear problems.

Pattern recognition using spiking neurons and firing rates

Different varieties of artificial neural networks have proved their power in several pattern recognition problems, particularly feed-forward neural networks. Nevertheless, these kinds of neural networks require of several neurons and layers in order to success when they are applied to solve non-linear problems. In this paper is shown how a spiking neuron can be applied to solve different linear and non-linear pattern recognition problems. A spiking neuron is stimulated during T ms with an input signal and fires when its membrane potential reaches a specific value generating an action potential (spike) or a train of spikes. Given a set of input patterns belonging to K classes, each input pattern is transformed into an input signal, then the spiking neuron is stimulated during T ms and finally the firing rate is computed. After adjusting the synaptic weights of the neuron model, we expect that input patterns belonging to the same class generate almost the same firing rate and input patterns belonging to different classes generate firing rates different enough to discriminate among the different classes. At last, a comparison between a feed-forward neural network and a spiking neuron is presented when they are applied to solve non-linear and real object recognition problems.

Evolving neural networks: a comparison between differential evolution and particle swarm optimization

Due to their efficiency and adaptability, bio-inspired algorithms have shown their usefulness in a wide range of different non-linear optimization problems. In this paper, we compare two ways of training an artificial neural network (ANN): Particle Swarm Optimization (PSO) and Differential Evolution (DE) algorithms. The main contribution of this paper is to show which of these two algorithms provides the best accuracy during the learning phase of an ANN. First of all, we explain how the ANN training phase could be seen as an optimization problem. Then, we explain how PSO and DE could be applied to find the best synaptic weights of the ANN. Finally, we perform a comparison between PSO and DE approaches when used to train an ANN applied to different non-linear problems.

Tuning the parameters of an integrate and fire neuron via a genetic algorithm for solving pattern recognition problems

Recently, it has been proven that spiking neurons can be used for some pattern recognition problems. Nonetheless, the spiking neurons models have many parameters that have to be manually adjusted in order to achieve the desired behavior. This paper has the purpose of showing an optimization method for one such model, the Integrate & Fire spiking model (I&F). A genetic algorithm (GA) is proposed to automatically adjust the parameters, removing the need of manual tuning and increasing efficiency. Initial experimentation is done by tuning the I&F model parameters by hand, to confirm the importance and relevance of determining the best parameter values. The GA is then used to automatically tune different parameter combinations of the pattern recognition model, which uses the I&F neuron as core, to determine which parameters are worth including in the GA. The proposed method was tested with five different datasets, where no change was required to apply the model to each. Very good results were achieved in all test cases, but experiments where parameters of the neuron model were included converged faster.

Classification of DNA Microarrays Using Artificial Bee Colony (ABC) Algorithm

DNA microarrays are a powerful technique in genetic science due to the possibility to analyze the gene expression level of millions of genes at the same time. Using this technique, it is possible to diagnose diseases, identify tumours, select the best treatment to resist illness, detect mutations and prognosis purpose. However, the main problem that arises when DNA microarrays are analyzed with computational intelligent techniques is that the number of genes is too big and the samples are too few. For these reason, it is necessary to apply pre-processing techniques to reduce the dimensionality of DNA microarrays. In this paper, we propose a methodology to select the best set of genes that allow classifying the disease class of a gene expression with a good accuracy using Artificial Bee Colony (ABC) algorithm and distance classifiers. The results are compared against Principal Component Analysis (PCA) technique and others from the literature.

3D Object Recognition Based on Some Aspects of the Infant Vision System and Associative Memory

A view-based method for 3D object recognition based on some biological aspects of infant vision is proposed in this paper. The biological hypotheses of this method are based on the role of the response to low frequencies at early stages as well as some conjectures concerning how an infant detects subtle features (stimulating points) from an object. In order to recognize an object from different images of it (at different orientations from 0° to 360°), we make use of a dynamic associative memory (DAM). As the infant vision responds to low frequencies of the signal, a low-filter is first used to remove high frequency components from the image. Then, we detect subtle features in the image by means of a random feature selection detector. At last, the DAM is fed with this information for training and recognition. To test the accuracy of the proposed model, we use the Columbia Object Image Library (COIL 100) database.

A computational approach for modeling the biological olfactory system during an odor discrimination task using spiking neuron

The biological olfactory system is capable of solving problems related to the olfactory information processing such as odor discrimination. This system is composed of three main parts: the layer receptors of the nose, the bulb and the piriform cortex [1]. The sense of smell is a chemical neural process where odorant molecules stimulate the olfactory system. These molecules are inhaled through the nose, where they contact the olfactory receptor neurons. The olfactory neurons transduce receptor activation into electrical signals in neurons. The signals travel along the olfactory nerve which terminates in the olfactory bulb. Finally, the olfactory bulb, which is composed of different cell layers, sends the information to the piriform cortex where discrimination between odors is performed. Odors are represented as patterns of neuronal activity. This representation may be encoded by space, time or a combination of both [2]. In this research, we described an approach for modeling the olfactory system in order to perform an odor discrimination task by means of the neural activity produced by a network of spiking neurons (SNs). Our model is composed of three layers. The first layer contains a set of neurons acting as receptive fields, which normalized the input stimulus and sent the information to the next layer. The second layer is modeled with a linear-type neuron, which received the information from each receptor neuron and enhances sensitivity to odor and discrimination through adjusting the synaptic connections. After that, obtained signal is directly injected to the third layer, which is composed of a network of SNs. During learning phase, synaptic connections are adjusted by means of an evolutionary learning approach [3]. Finally, discrimination between odors is performed by means of the neural activity recorded in the SNs. The described model, allow us to discriminate odors using the neural activity encoded by time or space. In the first case, SNs fire at similar firing rates with odors from the same class; on the other hand, odors from different classes provoke SNs fire at a firing rate different enough to discriminate among the odors. In the second case, a set of activated SNs determines an olfactory region which corresponds to the odor. To test the accuracy of the model during odor discrimination task a dataset was used [4]. This dataset contains 124 samples of two different classes of alcohol: butanol and ethanol. The dataset was split into training and testing dataset. During learning, the model performs with an accuracy of 96% and 92% in terms of the neuronal activity encoded by time and space, respectively. During testing, the model performs with 92% and 87%, respectively. The model learnt to discriminate odors by means of the neural activity encoded by time and space. Successful results suggest that the model could serve as a biological model to explain the process of odor discrimination in the olfactory system. Figure 1 Learning error and neural activity during an odor discrimination task. (a) – (b) Neural activity encoded by time. (c) – (d) Neural activity encoded by space.