Adelmo Luis Cechin

Real estate value at Porto Alegre city using artificial neural networks

Focuses on the MLP neural network in order to solve the problem of an apartments monetary worth appraisal in the Porto Alegre city (South Brazil). Many factors are involved in this calculation, like the size of the apartment, the environment conditions of the site, the actual conservation state of the apartment, the neighborhood, its geographical location in the city etc. Two databases were investigated: the first one is a list of apartments for sale and the second one is a list of apartments for rent. The analysis was performed with the use of both linear regression and neural network methods, with the purpose of comparison. The last one was used mainly to model the strong nonlinearities due to the geographical position of the apartments, since there is not a linear monotonic relation between position and value.

Cupin: A candidate molecular structure for the Nep1-like protein family

BackgroundNEP1-like proteins (NLPs) are a novel family of microbial elicitors of plant necrosis. Some NLPs induce a hypersensitive-like response in dicot plants though the basis for this response remains unclear. In addition, the spatial structure and the role of these highly conserved proteins are not known.ResultsWe predict a 3d-structure for the β-rich section of the NLPs based on alignments, prediction tools and molecular dynamics. We calculated a consensus sequence from 42 NLPs proteins, predicted its secondary structure and obtained a high quality alignment of this structure and conserved residues with the two Cupin superfamily motifs. The conserved sequence GHRHDWE and several common residues, especially some conserved histidines, in NLPs match closely the two cupin motifs. Besides other common residues shared by dicot Auxin-Binding Proteins (ABPs) and NLPs, an additional conserved histidine found in all dicot ABPs was also found in all NLPs at the same position.ConclusionWe propose that the necrosis inducing protein class belongs to the Cupin superfamily. Based on the 3d-structure, we are proposing some possible functions for the NLPs.

State automata extraction from recurrent neural nets using k-means and fuzzy clustering

This paper presents the use of a recurrent neural network to learn the dynamical behavior of the inverted pendulum and from this network to extract a finite state automata. Two clustering methods are compared for the automata extraction: the K-means method, and the construction of fuzzy membership functions. It is shown that the number of states for the fuzzy clustering method induces much less states than the K-means method.

Can Nep1-like proteins form oligomers?

Nep1-like proteins (NLPs) are a novel family of microbial elicitors of plant necrosis that induce a hypersensitive-like response in dicot plants. The spatial structure and role of these proteins are yet unknown. In a paper published in BMC Plant Biology (2008; 8:50) we have proposed that the core region of Nep1-like proteins (NLPs) belong to the Cupin superfamily. Based on what is known about the Cupin superfamily, in this addendum to the paper we discuss how NLPs could form oligomers. Addendum to: Cechin AL, Sinigaglia M, Lemke N, Echeverrigaray S, Cabrera OG, Pereira GAG, Mombach JCM. Cupin: A candidate molecular structure for the Nep1-like protein family. BMC Plant Biol 2008; 8:50.

Comparison of deterministic and fuzzy finite automata extraction methods from Jordan networks

This paper compares two methods for the extraction of finite state automata from recurrent neural networks (RNNs). Neural networks store the knowledge implicit in the data in their weights, but do not provide an easy explanation of this knowledge to the user. This is a difficult task due to the spatial (distributed information in the network) and temporal (network states) relations built by the network among the data. One form to present the knowledge stored inside a RNN is using finite state automata, which shows explicitly the relations among the variables and their temporal causality. In this paper, we treat the nonlinear dynamical system inverted pendulum and controller and compare the performance of the extraction algorithm using two clustering methods: k-means and fuzzy clustering in terms of exactness and knowledge conciseness.

The Protein Folding Problem Solved by a Fuzzy Inference System Extracted from an Artificial Neural Network

This paper reports the results of a rule extraction process from an Artificial Neural Network (ANN) used to predict the backbone dihedral angles of proteins based on physical-chemical attributes. By analyzing the fuzzy inference system extracted from the knowledge acquired by the ANN we want to scientifically explain part of the results obtained by the scientific community when processing the Hydropathy Index and the Isoeletric Point and also show that the rule extraction process from ANNs is an important tool that should be more frequently used. To obtain these results we defined a methodology that allowed us to formulate hypothesis statistically sustained and to conclude that these attributes are not enough to predict the backbone dihedral angles when processed by an ANN approach.

The interpretation of feedforward neural networks for secondary structure prediction using sugeno fuzzy rules

This paper presents results from the rule extraction process applied on multilayer neural networks for the prediction of protein backbone dihedral angles based on the physical-chemical attributes of the amino acids. The goals are to analyze the knowledge acquired by the neural network in the form of a fuzzy inference system of Sugeno fuzzy rules, to explain the results obtained by the scientific community when processing the Hydropathy Index and the Isoeletric Point and to show that the rule extraction process is an important tool to analyze neural networks. The proposed extraction algorithm and the rules are shown and discussed in the context of the application, relating them to the original proteins from which the data was computed. It is shown that the proposed rules are simple in the sense of being linear in relation to the input parameters, and therefore accessible to anyone with basic scientific knowledge. We also present a validation of rules from a statistical point of view.

Gene Expression Analysis using Markov Chains extracted from RNNs

This paper present a new approach for the analysis of gene exp res- sion, by extracting a Markov Chain from trained Recurrent Ne ural Networks (RNNs). A lot of microarray data is being generated, since ar ray technologies have been widely used to monitor simultaneously the express ion pattern of thou- sands of genes. Microarray data is highly specialized, invo lves several variables in which are complex to express and analyze. The challenge is to discover how to extract useful information from these data sets. So this w ork proposes the use of RNNs for data modeling, due to their ability to learn co mplex temporal non-linear data. Once a model is obtained for the data, it is p to ex- tract the acquired knowledge and to represent it through Mar kov Chains model. Markov Chains are easily visualized in the form of states gra phs, which show the influences among the gene expression levels and their cha nges in time.

Genetic algorithms to solve the power system restoration planning problem

This study reports the use of a Genetic Algorithm (GA) to solve the Power System Restoration Planning Problem (PSRP). The solution to the PSRP is described by a series of operations or a plan to be used by the Power System operator immediately on the occurrence of a blackout in the electrical power supply. Our GA uses new initialization and crossover operators based on the electrical power network, which are able to generate and maintain the plans feasible along GA runs. This releases the Power Flow program, which represents the most computer demanding component, from computing the fitness function of unfeasible individuals. The method was designed for large transmission systems and results for three different electrical power networks are shown: IEEE 14-Bus, IEEE 30-Bus, and a large realistic system.

Hybrid machine learning tools: INSS - A neuro-symbolic system for constructive machine learning

In this paper we present the INSS system, a new hybrid approach based upon the principles of KBANN networks. It represents an important improvement in comparison with its predecessor because the learning and the knowledge extraction process are faster and are accomplished in an incremental way . INSS offers a new approach applicable to constructive machine learning with high-performance tools, even in the presence of incomplete or erroneous data.