def learn_loop(dataset, answers, target): ### Just the digits from main dataset """Takes a dataset with expected values returns proper weights for each element of the vector space """ # set initial guess, initial weights, threshold, and learning rate binary_dataset = convert_dataset(dataset) [vector.append(1) for vector in binary_dataset] expected = [1 if answers[idx] == target else 0 for idx, x in enumerate(binary_dataset)] weights = [0 for elem in binary_dataset[0]] guesses = [0 for vector in binary_dataset] threshold = .5 l_rate = .05 ### Check each vector against expect and loop over again until 0 errors counter = 0 while True: for idy, vector in enumerate(binary_dataset): error = expected[idy] - guesses[idy] if dot_product(vector, weights) > threshold: #if true, set output to 1 guesses[idy] = 1 else: guesses[idy] = 0 weights = update_weights(weights, error, l_rate, vector) if expected != guesses and counter < 500: counter += 1 continue else: break return weights
def test_dot_product(self):
a = [0, 0, 0, 0, 4, 0, 0, 0]
b = [0, 0, 5, 0, 1, 0, 0, 0]
result = dp.dot_product(a, b)
self.assertEqual(result, 4)
def check_new_iris(vector, weights): """ Takes in a vector (representing a picture of a number) and weights Returns a guess as to it's value """ threshold = .5 computer_guesses = [] dot_product_list = [] for x in range(3): dot_product_list.append(dot_product(vector, weights[x])) if dot_product_list[x] > threshold: computer_guesses.append(x) if dot_product_list: return dot_product_list.index(max(dot_product_list)) else: return None
