Science and Technology
publicado em 08/07/2013 às 11h00:00
   Dê o seu voto:

Software capable of making predictions can aid in medical diagnosis

USP researchers and Unifesp were inspired by evolutionary theory of Charles Darwin to develop the computer program

font size

Some pieces of information on this page may have been automatically translated. Makernews is not responsible for the irregularities resulting from these translations. When in doubt,      consult the original text.

Researchers at the University of São Paulo (USP) in São Carlos, and the Universidade Federal de São Paulo (Unifesp) in São José dos Campos, were inspired by evolutionary theory of Charles Darwin to develop a program generator Decision Trees. The technique can, for example, assisting medical diagnosis or risk analysis of financial investments. Decision Trees are computational tools that give machines the ability to make predictions based on analysis of historical data.

We have developed an evolutionary algorithm, ie, that mimics the process of human evolution to generate solutions, said Rodrigo Barros Coelho, PhD Laboratory of Computer Bioinspirada (bio) ICMC and FAPESP fellowship.

The evolutionary computation, Barros said, is one of several techniques bioinspiradas, ie in nature seeking solutions to computational problems. It is remarkable how nature is extremely complicated solutions to problems. There is no doubt that we need to learn from it, said Barros.

According to Barros, the software developed in his PhD is able to automatically create programs generators Decision Trees. For this, makes random mating between the existing program code generating children.

These children may eventually mutate and evolve. After a while, it is expected that generation programs Evolved Decision Trees are getting better and our algorithm selects the best of all, Barros said.

But while the process of natural selection in the human species takes hundreds or even thousands of years, the computation lasts only a few hours, depending on the problem to be solved. We set out as a hundred generations of the evolutionary process, said Barros.

<b> Artificial Intelligence </ b>

In Computer Science, is called heuristic ability of a system to develop innovations and techniques to achieve a particular purpose.

The software developed by Barros fall within the area of ​​hyper-heuristics, recent topic in the field of evolutionary computation that aims automatic generation of heuristics tailored to a particular application or set of applications.

It is a preliminary step toward the ultimate goal of artificial intelligence: to create machines capable of developing solutions to problems without being explicitly programmed to do so, detailed Barros.

The work led to the article Hyper-Heuristic Evolutionary Algorithm for Automatically Designing Decision-Tree Algorithms, awarded in three categories in the Genetic and Evolutionary Computation Conference (Gecco), the largest event in the field of evolutionary computation in the world, held in July in Philadelphia USA.

Besides Barros, also are authors of the article professors André Carlos Ponce Leon Ferreira de Carvalho, the head researcher at the ICMC, Márcio Port Basgalupp, Federal University of São Paulo (Unifesp), and Alex Freitas, University of Kent at UK, which took over the co-orientation.

<b> Recognition </ b>

The results of the winning article are also a result of the research project "Genetic Programming to the evolution of algorithms for induction of decision trees", conducted by Basgalupp supported by FAPESP through the Program for Young Researchers in Emerging Centers.

The authors were invited to submit the article for the journal Evolutionary Computation Journal, published by the Massachusetts Institute of Technology (MIT). Work still undergo review but, as was submitted by invitation, has a good chance of being accepted, said Barros.

The research, to be completed only in 2013, also gave an invited paper published in the Journal of the Brazilian Computer Society, after being elected as the best work at the National Conference on Artificial Intelligence, 2011.

Another paper, presented at the 11th International Conference on Intelligent Systems Design and Applications, held in Spain in 2011, yielded invitation for publication in Neurocomputing.

<i> information Fapesp </ i>

  • Share this pageShare this page
  • Share this pageCorrect
  • ShareShare
  • AlertAlert
Reduced link: 
  • You are recommending this story: Software capable of making predictions can aid in medical diagnosis
  • Fill in the following form to send your recommendation to your friend:

  • You are suggesting a correction for this story: Software capable of making predictions can aid in medical diagnosis

Receba notícias do iSaúde no seu e-mail de acordo com os assuntos de seu interesse.
Seu nome:
Seu email:
Desejo receber um alerta com estes assuntos:
Evolutionary computation    forecasting    medical diagnosis    Computer Bioinspirada    Charles Darwin   
Leave your comment
(Required fields are marked with an *)

(Your email address will never be published or shared.)

Enter the letters and numbers below and click in the button "send"

  • Twitter iSaúde
Informe Saúde printed version

Recommend the portal
Close [X]
  • You are recommending this story:
  • Fill in the following form to send your recommendation to your friend:

RSS news from the portal  iSaú
Get the newsletter of the portal  iSaú
Recommend the portal iSaú
News from  iSaú in your blog or website.
Get news on the subject of your interest.
© 2000-2011 Todos os direitos reservados.