backward.py

for i in range(train.shape[0]):
        v=[]
        y=[]
        inputNeuron = [train[i]]
        v.append(inputNeuron)
        y.append(inputNeuron)
        
        for r in range(len(layerNeurons) -1):
            w=weight[r]
            matout=np.matmul(inputNeuron,w)
            v.append(matout)
            matout=sigmoid(matout)
            y.append(matout)
            inputNeuron=matout
        
        
        lastY=len(y)-1
        errs = 0.5 * (y[lastY] - train_Y[i])*(y[lastY] - train_Y[i])
        if errs.sum() < min_err:
            min_err = errs.sum()
            best_w = weight
        
        delta =[]
        d=(y[lastY] - train_Y[i])* derivative(y[lastY])
        delta.append(d)
        #print(delta[0])
        
        for r in range(len(layerNeurons)-2,0,-1):
            d= (np.matmul(weight[r],delta[len(layerNeurons)-r-2].T)).T
            d= d*derivative(y[r])
            delta.append(d)
        
        
        delta.reverse()
        
        delw=[]
        for i in range(len(layerNeurons)-1):
            w = np.random.uniform(0,0,(layerNeurons[i],layerNeurons[i+1]))
            delw.append(w)
        
        for r in range(len(delw)-1,0,-1):
            delw[r]=np.matmul(np.array(y[r]).T,delta[r])