Papers by Ronaldo Trindade
Anais do 9. Congresso Brasileiro de Redes Neurais, 2016
This paper presents the implementation of a neural network based on a PIC microcontroller. We use... more This paper presents the implementation of a neural network based on a PIC microcontroller. We use the digital inputs of the PIC16F877A microcontroller as an input layer. The hidden and the output layers were implemented by software considering the limitations of the microcontroller with respect to the memory use for the floating point operations. The Output layer was directly coupled to the digital outputs of the microcontroller which in turn were conected to LEDs. For validating the project, we used a 3x3 matrix for representing the vowels. The system responds to this input by turning on the corresponding LED, at the output.
labplan.ufsc.br
This work presents a linearization and self-calibration system for nonlinear sensors. The impleme... more This work presents a linearization and self-calibration system for nonlinear sensors. The implemented system is composed of hardware and software solutions. The hardware solution involves signal conditioning circuitry and a microcontroller. The software solution comprises radial basis function neural networks with multi-objective learning. The output linear weights are determined by means of the multi-objective least square technique. By varying the width of the radial basis functions, different Pareto solutions are obtained, in order to find the network with the most adequate structure for the problem. The advantage of using multi-objective learning in the context of sensor linearization is to determine an adequate network for the problem that presents low reduced structural complexity, thus reducing the hardware implementation cost.
… em condições de incêndio, Jan 1, 2003
Ao meu pai Sérgio e ao meu avô Eraldo.
Uploads
Papers by Ronaldo Trindade