새싹 AI데이터엔지니어 핀테커스 4주차 (목) - ML 관련 수학 (3)

2023-09-21 19th Class

❤️ 배운 것

머신러닝 관련 수학 공식 파이썬 알고리즘으로 구현하기 - 3
(2차원 list 편)

• 정확도

def e177():  
    # accuracy  
    predictions = [0, 1, 0, 2, 1, 2, 0]  
    labels = [1, 1, 0, 0, 1, 2, 1]  
    n_correct = 0  

    for pred_idx in range(len(predictions)):  
        if predictions[pred_idx] == labels[pred_idx]:  
            n_correct += 1  


    accuracy = n_correct / len(predictions)  
    print("accuracy[%]: ", accuracy*100, '%')

accuracy[%]:  57.14285714285714 %

• confusion vector (confusion matrix 간소화)

def e178():  
    # confusion vector  
    predictions = [0, 1, 0, 2, 1, 2, 0]  
    labels = [1, 1, 0, 0, 1, 2, 1]  

    n_classes = None  
    for label in labels:  
        if n_classes == None or label > n_classes:  
            n_classes = label  

    n_classes += 1  

    class_cnts, correct_cnts, confusion_vec = list(), list(), list()  
    for _ in range(n_classes):  
        class_cnts.append(0)  
        correct_cnts.append(0)  
        confusion_vec.append(None)  

    for pred_idx in range(len(predictions)):  
        pred = predictions[pred_idx]  
        label = labels[pred_idx]  

        class_cnts[label] += 1  
        if pred == label:  
            correct_cnts[label] += 1  

    for class_idx in range(n_classes):  
        confusion_vec[class_idx] = correct_cnts[class_idx]/class_cnts[class_idx]  

    print("confusion vector: ", confusion_vec)

confusion vector:  [0.5, 0.5, 1.0]

• histogram

def e179():  
    # histogram  
    scores = [50, 20, 30, 40, 10, 50, 70, 80, 90, 20, 30]  
    cutoffs = [0, 20, 40, 60, 80]  
    histogram = [0, 0, 0, 0, 0]  

    for score in scores:  
        if score > cutoffs[4]:  
            histogram[4] += 1  
        elif score > cutoffs[3]:  
            histogram[3] += 1  
        elif score > cutoffs[2]:  
            histogram[2] += 1  

        elif score > cutoffs[1]:  
            histogram[1] += 1  
        elif score > cutoffs[0]:  
            histogram[0] += 1  
        else:  
            pass  

    print("histogram of the scores: ", histogram)

histogram of the scores:  [3, 3, 2, 2, 1]

• abs

def e180():  
    # abs  
    numbers = [-2, 2, -1, 3, -4, 9]  
    abs_numbers = list()  

    for num in numbers:  
        if num < 0:  
            abs_numbers.append(-num)  
        else:  
            abs_numbers.append(num)  

    print(abs_numbers)

[2, 2, 1, 3, 4, 9]

• manhattan distance

def e181():  
    # manhattan distance  
    v1 = [1, 3, 5, 2, 1, 5, 2]  
    v2 = [2, 3, 1, 5, 2, 1, 3]  

    m_distance = 0  
    for dim_idx in range(len(v1)):  
        sub = v1[dim_idx] - v2[dim_idx]  
        if sub < 0:  
            m_distance += -sub  

        else:  
            m_distance += sub  
    print("manhattan distance: ", m_distance)

manhattan distance:  14

• nested list

def e182():  
    # nested list  
    scores = [[10, 20, 30], [50, 60, 70]]  
    print(scores)  
    print(scores[0])  
    print(scores[1])  
    print(scores[0][0], scores[0][1], scores[0][2])  
    print(scores[1][0], scores[1][1], scores[1][2])

[[10, 20, 30], [50, 60, 70]]
[10, 20, 30]
[50, 60, 70]
10 20 30
50 60 70

• nested list 원소 접근

def e183():  
    # nested list 원소 접근  
    scores = [[10, 20, 30], [50, 60, 70]]  

    for student_scores in scores:  
        print(student_scores)  
        for score in student_scores:  
            print(score)

[10, 20, 30]
10
20
30
[50, 60, 70]
50
60
70

• 2차원 list 열/행의 평균 구하기

def e184():  
    # 학생별 평균점수 구하기  
    scores = [[10, 15, 20], [20, 25, 30], [30, 35, 40], [40, 45, 50]]  

    n_class = len(scores[0])  
    student_score_means = list()  

    for student in scores:  
        score_sum = 0  
        for score in student:  
            score_sum += score  

        student_score_means.append(score_sum / n_class)  

    print("mean of students' scores: ", student_score_means)

mean of students' scores:  [15.0, 25.0, 35.0, 45.0]

• 2차원 list sum, mean

def e185(scores=None):  
    # 과목별 평균  
    if scores is None:  
        scores = [[10, 15, 20], [20, 25, 30], [30, 35, 40], [40, 45, 50]]  

    col_sums = [0 for x in range(len(scores[0]))]  

    n_cols = len(scores[0])  
    rows = len(scores)  

    for i in range(n_cols):  
        for j in range(rows):  
            col_sums[i] += scores[j][i]  

    col_means = [x/rows for x in col_sums]  

    print("sum of classes' scores: ", col_sums)  
    print("mean of classes' scores: ", col_means)  

    return col_means

sum of classes' scores:  [100, 120, 140]
mean of classes' scores:  [25.0, 30.0, 35.0]

• 2차원 list mean subtraction

def e186():  
    def e185_1(scores=None):  
        # 과목별 평균  
        if scores is None:  
            scores = [[10, 15, 20], [20, 25, 30], [30, 35, 40], [40, 45, 50]]  

        col_sums = [0 for x in range(len(scores[0]))]  

        n_cols = len(scores[0])  
        rows = len(scores)  

        for i in range(n_cols):  
            for j in range(rows):  
                col_sums[i] += scores[j][i]  

        col_means = [x / rows for x in col_sums]  

        print("sum of classes' scores: ", col_sums)  
        print("mean of classes' scores: ", col_means)  

        return n_cols, rows, col_sums, col_means  

    # 과목별 평균 mean subtraction    scores = [[10, 15, 20], [20, 25, 30], [30, 35, 40], [40, 45, 50]]  

    n_cols, rows, col_sums, col_means = e185_1(scores)  

    scores_ms = list()  
    for idx, student in enumerate(scores):  
        m_subtracted = [x-col_means[i] for i, x in enumerate(student)]  
        scores_ms.append(m_subtracted)  

    print("="*20)  
    print("mean subtracted scores: ", scores_ms)  

    _, _, col_sums_ms, col_means_ms = e185_1(scores_ms)

sum of classes' scores:  [100, 120, 140]
mean of classes' scores:  [25.0, 30.0, 35.0]
====================
mean subtracted scores:  [[-15.0, -15.0, -15.0], [-5.0, -5.0, -5.0], [5.0, 5.0, 5.0], [15.0, 15.0, 15.0]]
sum of classes' scores:  [0.0, 0.0, 0.0]
mean of classes' scores:  [0.0, 0.0, 0.0]

• 2차원 list variace, standard deviation

def e187():  
    def e185_2(scores=None):  
        # 과목별 평균  
        if scores is None:  
            scores = [[10, 15, 20], [20, 25, 30], [30, 35, 40], [40, 45, 50]]  

        col_sums = [0 for x in range(len(scores[0]))]  

        n_cols = len(scores[0])  
        rows = len(scores)  

        for i in range(n_cols):  
            for j in range(rows):  
                col_sums[i] += scores[j][i]  

        col_means = [x / rows for x in col_sums]  

        # print("sum of classes' scores: ", col_sums)  
        # print("mean of classes' scores: ", col_means)  
        return n_cols, rows, col_sums, col_means  

    #################  
    #  분산 표준편차  #    #################    scores = [[10, 15, 20], [20, 25, 30], [30, 35, 40], [40, 45, 50]]  

    # MOS  
    squared = [[score**2 for score in student] for student in scores]  
    _, _, _, moss = e185_2(squared)  

    # SOM  
    _, _, _, col_means = e185_2(scores)  
    soms = [x**2 for x in col_means]  

    variances = [m - s for m, s in zip(moss, soms)]  
    stds = [v**0.5 for v in variances]  

    print("variance values: ", variances)  
    print("std values: ", stds)

variance values:  [125.0, 125.0, 125.0]
std values:  [11.180339887498949, 11.180339887498949, 11.180339887498949]

• 2차원 list standardization

처음에는 함수로만 작성했지만 여기부터는 클래스로 작성

class MathBase:  
    def __init__(self, scores=None):  
        if scores is None:  
            self.scores = [[10, 15, 20], [20, 25, 30], [30, 35, 40], [40, 45, 50]]  
        else:  
            self.scores = scores  

        self.standardized = self.standardize(self.scores)  

    def get_sum_mean(self, scores=None):  
        if scores is None:  
            scores = self.scores  

        col_sums = [0 for x in range(len(scores[0]))]  

        n_cols = len(scores[0])  
        rows = len(scores)  

        for i in range(n_cols):  
            for j in range(rows):  
                col_sums[i] += scores[j][i]  

        col_means = [x / rows for x in col_sums]  

        return n_cols, rows, col_sums, col_means  

    def get_var_std(self, scores=None):  
        if scores is None:  
            scores = self.scores  
        # MOS  
        squared = [[score ** 2 for score in student] for student in scores]  
        _, _, _, moss = self.get_sum_mean(squared)  

        # SOM  
        _, _, _, col_means = self.get_sum_mean(scores)  
        soms = [x ** 2 for x in col_means]  

        variances = [m - s for m, s in zip(moss, soms)]  
        stds = [v ** 0.5 for v in variances]  

        return variances, stds  

    def standardize(self, scores=None):  
        if scores is None:  
            scores = self.scores  

        _, _, _, col_means = self.get_sum_mean(scores)  
        _, stds = self.get_var_std(scores)  

        standardized = [[(score - col_means[idx]) / stds[idx] for idx, score in enumerate(student)]  
                        for student in scores]  

        return standardized  


def e188():  
    # standardization  
    scores = [[10, 15, 20], [20, 25, 30], [30, 35, 40], [40, 45, 50]]  
    base = MathBase(scores)  
    standardized = base.standardize()  
    print("standardized scores: ", standardized)  

    var, std = base.get_var_std(standardized)  
    print("vars and stds after standardization: ", var, std)

standardized scores:  [[-1.3416407864998738, -1.3416407864998738, -1.3416407864998738], [-0.4472135954999579, -0.4472135954999579, -0.4472135954999579], [0.4472135954999579, 0.4472135954999579, 0.4472135954999579], [1.3416407864998738, 1.3416407864998738, 1.3416407864998738]]
vars and stds after standardization:  [1.0, 1.0, 1.0] [1.0, 1.0, 1.0]

💛 배운점/느낀점

• 오늘은 (가독성과 별개일 수 있지만..) list comprehension을 이중으로 작성하는것에 익숙해지려고 최대한 많이 작성해봤다.
• 평상시랑 다르게 nested list를 라인디버깅을 최소화하면서 암산(?)으로 해보려고 노력했다.