A research on evolutionary automatic construction of image recognition procedure based on machine learning

在本课题中，我们研究了基于进化计算的图像识别器的自动构造。因此，我们提出了一种称为“进化决策网络”的方法。在该系统中，通过向计算机提供学习图像，自动生成将输入信号及其参考信号分类为多个类别的网络。网络的每个节点根据其特征值的大小对信号进行分支。我们进行了包括医学图像识别在内的多个实验，验证了该方法的有效性。

In this project, we researched automatic construction of image recognizer by using evolutionary computation. As the result, we developed a method named "Evolutionary Decision Network". In this system, by giving learning images to computer, a network that classifies the input signal and its referential signals into several classes is automatically generated. Each node of the network branches the signal according to the size of its feature value. We performed several experiments including medical image recognition, and verified the usefulness of the proposed method.