Deep Neural Networks applied to medical Diagnostic
We have developed a methodology to train Deep Neural Networks (DNN) whose operative structure encloses short term memory elements in the form of a space-time discharge matrix. Here highly compressed data is orderly released, creating a new space-expanded picture of the chosen object. We prove that this expanding picture is a rich source of features of the object and that a second shallow network can be quickly trained to separate the object from white noise. The final result is robotic eye whose six layer neural processor have a very slender figure and generates the capacity to vigorously track and recognize the chosen object when it moves freely in a 3D space and is captioned in real time with a regular web can.
We incorporate new deep learning methods and control mechanisms in order to build an artificial vision system capable of exploring Pap smear databases and automatically learn about cell classification. To promote hierarchical data features our deep architecture is organized in two operative levels, controlled by independent neural agents. For this study the quality of being and agent implies the capacity to perform a useful job without external intervention and satisfying the four weak conditions of: Autonomy, Social ability, Reactivity and Pro-activenes
Artificial Vision using neural networks
We have studied and developed the behavior of two specific neural processes,
used for vehicle driving and path planning, in order to control mobile robots.
Each processor is an independent agent defined by a neural network trained for a defined task. Through simulated evolution fully trained agents are encouraged to socialize by opening low bandwidth, asynchronous channels between them. Under evolutive pressure agents spontaneously develop communication skills (protolanguage) that take advantages o