The perceptron is an algorithm for learning a binary classifier called a threshold function: a function that maps its input x (a real-valued vector) to an output value f(x) (a single binary value).
The perceptron is implemented as follows:
Inputs
- 10 'x' vectors (8 bits);
- 10 'w' vectors (9 bits);
- bias vector (9 bits).
All the inputs are considered in the range [-1,1] with the standard 2’s complement notation.
The activation function is a sigmoid.
Output
- vector of 16 bits with value in range [0,1].
The behaviour and structure of the perceptron is described with the following language:
- VHDL
The tests to verify the correctness of the system were made with:
- VHDL (for the writing of testbenches)
- ModelSim (for the simulation phase)
- Python (for the check of the results)
The synthesis and analysis phases were carried out using the following tool:
- Xilinx Vivado