simple_som.py

""" Dataset """

v1 = [1, 1, 0, 0]
v2 = [1, 0, 0, 0]
v3 = [0, 0, 0, 1]
v4 = [0, 0, 1, 1]

inputs = [v1, v2, v3, v4]
dataset_length = len(inputs)

print(dataset_length)


""" Training """

input_length = len(v1) # length of 1 input vector

weights0 = [0.09, 0.2, 0.5, 0.95]
weights1 = [0.96, 0.15, 0.9, 0.35]

weights_length = len(weights0)

iteration_number = 100

for i in range(0,iteration_number):
    if i < 10:
        learning_rate = 0.6;
    else: 
        learning_rate = 0.3;
    
    for k in range(0,dataset_length): # for 1 iteration, we use each input vector one by one  
        distance0 = 0
        distance1 = 0
        for j in range(0,input_length):
            distance0 += pow(inputs[k][j] - weights0[j],2)
            distance1 += pow(inputs[k][j] - weights1[j], 2)
        distances = [distance0, distance1];
        dmin = min(distances)
        index = distances.index(min(distances))
        
        if index == 0: #update w0
            for j in range(0,weights_length):    
               weights0[k] = weights0[k] + learning_rate*  (inputs[k][j] - weights0[k])
        else: #update w1
            for j in range(0,weights_length):    
               weights1[k] = weights1[k] + learning_rate*  (inputs[k][j] - weights1[k])   
            
    
    print(weights0)  
    print(weights1)  
        
""" Testing """

test_v1 = [0, 0, 1, 0.9]
test_v2 =[0, 0, 0.8, 0.9]
test_v3 =[0.7, 0, 0, 0]
test_v4 =[0.7, 0.9, 0, 0]

test_inputs = [test_v1, test_v2, test_v3, test_v4]
test_dataset_length = len(test_inputs)

for k in range(0,test_dataset_length):
    distance0 = 0
    distance1 = 0
    for j in range(0,input_length):
        distance0 += pow(test_inputs[k][j] - weights0[j],2)
        distance1 += pow(test_inputs[k][j] - weights1[j], 2)
        distances = [distance0, distance1];
        dmin = min(distances)
        index = distances.index(min(distances))
        
    print(test_inputs[k], "vector belongs to", index, ". class")