J.D.S. da Silva

Comparison of Artificial Neural Network and Regression Models in Software Effort Estimation

Good practices in software project management are basic requirements for companies to stay in the market, because the effective project management leads to improvements in product quality and cost reduction. Fundamental measurements are the prediction of size, effort, resources, cost and time spent in the software development process. In this paper, predictive Artificial Neural Network (ANN) and Regression based models are investigated, aiming at establishing simple estimation methods alternatives. The results presented in this paper compare the performance of both methods and show that artificial neural networks are effective in effort estimation.

A neural network application for attack detection in computer networks

This work presents a network intrusion detection method, created to identify and classify illegitimate information in TCP/IP packet payload based on the Snort signature set that represents possible attacks to a network. For this development, a type of neural network named Hamming net was used. The choice of this network is based on the interest to investigate its adequacy to classify network events in real-time, due to its capability to learn faster than other neural network models, such as, multilayer perceptrons with backpropagation and Kohonen maps. A Hamming net does not require exhaustive training to learn. TCP/IP packet payloads were used as input pattern to the Hamming net and Snort signature as exemplar patterns. The challenges faced in modeling the input and exemplar data and the strategies adopted to capture and scan relevant data in TCP/IP packets and in Snort signatures are described in this paper. In addition, the application architecture, the processing stages and some test results are presented.

Fuzzy Decision in Airplane Speed Control

A fuzzy decision system for helping air-traffic experts in controlling airplane velocities and in keeping an airplane flight within several constraints established to air lane sections is proposed in this paper. Automatic systems for air-traffic control are essential due to the ever increasing number of airplanes flying all over the world, the amount of environmental and airplane constraints and the necessity to guarantee the safety both for flights and for air-traffic control operators. The proposed system uses Mamdani direct inference method. Results show the effectiveness of the developed fuzzy system in controlling the airplane velocity to achieve the desired performance and encourage the adequacy of the system to include several different variables usually employed in air-traffic control.

Image based autonomous navigation with fuzzy logic control

This work presents a computational model of an adaptive image based autonomous navigation system using fuzzy logic control. A mobile robot moves in an environment from which images provide the necessary guidance information. The fuzzy logic system determines the directions to be followed for each point on the road. The system shows autonomy in adaptively navigating in the environment with image feedback, and automatically correcting its trajectory based on the information from the images. This is analog to human behavior while navigating in an environment. Experiments were conducted in an indoor environment in which simple computer vision techniques were applied to extract features of the road, and to feed the fuzzy logic system with linguistic variables that carried information about the directions of the road. The system decides the direction to follow from the actual position. The rules were constructed by considering a common sense driver model. The adequacy of the model for robot image based navigation is shown.

Fuzzy reference gain-scheduling approach as intelligent agents: FRGS agent

Goal driven intelligent agents and fuzzy reference gain-scheduling (FRGS) approach are described As interchangeable concepts that are able to deal with dynamic complex problems. It is advocated that the FRGS approach may be viewed as an autonomous goal-based agent, that is, a fuzzy logic based agent able to autonomously adapt itself to environmental changes introduced by external inputs. The concept of fuzzy systems and intelligent agent are employed in decision-making problems to lead to a certain degree of autonomy in decision support system. Although the FRGS method was originally proposed for control application, this approach was extended to decision-making tasks due to its ability of emulating human reasoning. This new agent approach uses the external input information also denominated reference (goal) as the driven mechanism to determine the behavior of the system in order to achieve the desired objectives (goal). Thus, the FRGS approach can be modeled in the framework of an adaptive and goal (also context or environment) driven agent.

The island model parallel GA and uncertainty reasoning in the correspondence problem

This work presents a pointwise approach to the correspondence problem in computer vision using contextual and structural features of a point. Multiple points are simultaneously considered, under the structural coherence constraint related to the similarity of the geometric features of emerging polygonal regions. Perceptron neural networks compute structural features. The correspondence is achieved by optimizing similarity measurements among the points features and by satisfying the structural coherence constraint. The island model parallel genetic algorithm (GA) searches a very large space by evolving several populations separately, using the Dempster-Shafer calculus for fitness evaluation. Changes were made in the parallel GA model in an attempt to introduce a higher level of diversity in the process. The occlusion problem is approached by iterating the whole process, alternating the reference image. Experimental results using a pair of real world indoor images demonstrate the usefulness of the approach for the correspondence problem. The reported simulations were conducted using 9 virtual machines. Comparisons made with previous work show a higher accuracy for the maximization process. A disparity map is constructed by considering the set of corresponding points as control points and by minimizing the differences in similarity among the image points.

Restoring images with a multiscale neural network based technique

This paper describes a neural network based multiscale image restoration approach using multilayer perceptron neural networks trained with artificially degraded images of gray level co-centered circles. The main goal of the approach is to make the neural network learn inherent space relations of the degraded pixels in restoring the image. In the conducted experiment, the degradation is simulated by filtering the image with a low pass Gaussian filter and adding noise to the pixels at preestablished rates. Degraded image pixels make the input and nondegraded image pixels make the target output for the supervised learning process. The neural network performs an inverse operation by recovering a quasi-non-degraded image in terms of least squared. The main difference of the present approach to existing ones relies on the fact the space relations are taken from different scales, thus providing correlated space data to the neural network. The approach attempts to develop a simple method that provide good restored versions of degraded images, without the need of a priori knowledge or estimation of the possible image degradation causes. The multiscale operation is simulated by considering different window sizes around a pixel. In the generalization phase the neural network is exposed to indoor, outdoor, and satellite degraded images following the same steps used to degrade the artificial image of circles. The neural network restoration results show the proposed approach is promising and may be used in restoration processes with the advantage it does not need a priori knowledge of the degradation causes.

Multiple correspondences in stereo vision under a genetic algorithm approach

This paper presents a multiple point hierarchical approach to the stereo correspondence problem in computer vision. The low-level processing employs an area-and-token hybrid method to obtain, for distinctive points in one image, a set of points in the other image that are candidates for correspondence. The refinement of the set of low-level correspondences obtained is performed by a high-level N-point simultaneous correspondence process, a new constraint introduced for this problem. The high-level processing uses a genetic algorithm approach for searching the solution space. Experimental results show the effectiveness of the method on real world scenes.

Integration of hierarchical censored production rule (HCPR)-based system and neural networks

Over the past few years, researchers have successfully developed a number of systems that combine the strength of the symbolic and connectionist approaches to artificial intelligence. Most of the efforts have employed standard production rules, IF THEN as underlying symbolic representation. This paper is an attempt towards integrating hierarchical censored production rule based system and neural networks. A HCPR has the form: decision (if, precondition) (unless, censor conditions) (generality, general information) (specificity, specific information) which can be made to exhibit variable precision in the reasoning such that both certainty of belief in a conclusion and its specificity may be controlled by the reasoning process. The proposed hybrid system would have numerous applications where decision must be taken in real time and with uncertain information.

Atmospheric temperature retrieval using a Radial Basis Function Neural Network

Vertical temperature profiles are obtained from measured satellite radiance data by using a Radial Basis Function Neural Network (RBF-NN). The RBF-NN is trained with data provided by the direct model, characterised by the Radiative Transfer Equation. The results are compared with regularisation-based inverse solutions. The approach is tested using satellite radiances, and the inversion temperature profile is compared with radiosonde temperature measurements. Analysis reveals that the generated profiles are closely approximate to previous results, showing the methodology adequacy. ANNs are useful because of the parallelism and implementation simplicity, turn hardware implementation possible, that may imply in on-board and real-time systems.