neuron.py

class Neuron:
    def __init__(self,input_size, learning_rate):
        self.w = np.random.random((input_size,1))-0.5  # self.w is a 2 dimensional column vector
        self.b  = np.random.random(1)-0.5
        self.learning_rate = learning_rate
    #forward pass
    def forward(self,x):
        a = x.T.dot(self.w) + self.b
        return a
    def loss(self,x,y):
        a = self.forward(x)
        #Mean squared error loss
        L = (y-a)**2
        return L
    #backpropagation
    def backward(self,x, y, y_hat):
        dw = 2.*(y_hat - y) * x.T
        db = 2.*(y_hat - y) * 1.
        return dw,db
    #training
    def train(self,x,y):
        #ensuring, that x is a 2 dimensional column vector.
        x = x.reshape(-1,1)
        a = self.forward(x)
        dw,db = self.backward(x,y,a)
        #gradient descent
        self.w = self.w - self.learning_rate*dw.T
        self.b = self.b - self.learning_rate*db.T