I added 3 output neurons meaning the neural network will respond in tems of class 1: 0.01, class 2: 0.03, class 3: 0.96. To implement that I had to add softmax fucntion chnage my forwardPass and backwardPass implementations to work with new matrixes.