Context
The goal is to train an AI model to predict whether a cancer is malignant or benign on a dataset of breast cancer diagnosis in the Wisconsin.
Dataset describe the (30) characteristics of a cell nucleus of breast mass extracted with fine-needle aspiration (FNA) on 569 patients.
Neuron network
input
30
layer 1
16
sigmoid
layer 2
16
sigmoid
output
2
softmax
34 neurons
Hyper parameters
x 0.75
learning rate
x 0.10
final learning rate
500
epochs
50 %
train/test dataset ratio
Training
The training phase initialize a model with random attributes.
The model is tested
(fowrward pass) on the shuffled patients of the training dataset. By comparaing the results of
each patients with the expected ones, an error is computed. Attributes of the model are
corrected (error back-propagation) to fit a little more with these expected results (backward
pass) according to these errors. This process is repeated x time (epochs).
Prediction
The prediction use the trained model.
The model is tested (fowrward pass) on the shuffled
patients of the testing dataset. By comparaing the results of each patients with the expected
ones.
Global accuracy:
0.00 %