You are required to use machine learning for fault detection in a robotic arm application. Two features obtained from sensor data were identified as the most discriminant. The diagram below shows the training data plotted in the feature space. The triangles correspond to fault condition and circles to normal condition. The rectangle with a question mark as a new sample we received, and we want to classify it into fault or normal. If you are using a Multi Layer Perceptorn to solve the problem presented, how many input neurons and output neurons you would use?

Feature 1