Pure Python Neural Network Function for Classification or Regression Problems. The function returns: 1) the final weigths, 2) the hidden layers activation function and 3) the output activation function
- Trainning: Full Batch.
- Activation Functions: "sigmoid", "tanh", "identity", "relu" or "gaussian".
- Output Activation Functions: "sigmoid", "tanh", "identity", "relu", or "gaussian".
- Epochs: The total number of iterations.
- Layers: List with the number of hidden layers and quantity of neurons. Ex: layers[0] = Perceptron; layers[1, 7] = Two hidden layers, the first one with 1 neuron and the second one with 7 neurons.
- Loss Function: "bin_cross_entropy" (Binary Output) or "mse" (Linear Output).
- For the Binary Output case a confusion matrix table and a loss function graph with the accuracy (acc) for each iteration is provided.
- For the Linear Output case a loss function graph for each iteration is provided.
- Finnaly a prediction function for the Binary Output (predict_bin_nn) and for the Linear Output (predict_nn) is also included.